Language variation - European perspectives VII : Selected papers from the Ninth International Conference on Language Variation in Europe (ICLaVE 9), Malaga, June 2017 9789027262073, 9027262071

Table of contents :
Intro
Language Variation --
European Perspectives VII
Editorial page
Title page
Copyright page
Table of contents
Preface
Local committee
Introduction
Scope of the volume
Overview of the chapters
Chapter 1. Language hybridism: On the origin of interdialectal forms
1. Introduction
2. Theoretical aspects: The origin of interdialectal forms
2.1 The neutrality hypothesis
2.2 The imperfect learning hypothesis
2.3 The interdialectal forms in Canarian Spanish
3. Analysing interdialectal forms
3.1 Linguistic variables
3.2 Phonetic distance between vernacular and standard forms 3.3 Individual variation3.4 The role of attitudes
4. Conclusions
References
Chapter 2. Of clocks, clouds and sound change
1. Introduction
2. The Neogrammarian legacy
3. Formal theory: Generative phonology and Optimality Theory
3.1 The life cycle of sound change
3.2 Awareness
4. Sociolinguistics
4.1 Exceptionlessness versus lexical diffuseness in the sociolinguistic study of sound change
5. Cognitivist approaches
6. Towards an integrated theory
References
Chapter 3. Evaluations of foreign accent in a purist speech community: The case of Iceland
1. Introduction 2. Evaluating an accent3. The language situation in Iceland
4. Method
4.1 Speakers
4.2 Procedure
4.3 Survey and data analysis
5. Results
5.1 Evaluations of personality traits
5.2 Gender
5.3 Age
6. Discussion
7. Conclusion
References
Chapter 4. C'era i fascisti e i tedeschi: Instances of linguistic simplification in a corpus of Italiano popolare
1. Introduction
2. On italiano popolare
3. Linguistic simplification in italiano popolare
4. Research questions and methodology
5. Analysis
5.1 New features
5.2 Case study: Existential constructions
6. Conclusions
Abbreviations AcknowledgmentsReferences
Chapter 5. Language change caught in the act: A case study of Frisian relative pronouns
1. Introduction
2. Omrop Fryslân
3. Frisian in the Netherlands
4. Relative pronouns
5. Method
5.1 The data
5.2 Procedure
5.3 Analysis
6. Results
7. Discussion
8. Conclusion
References
Chapter 6. Virtual sociolinguistics: From real-time surveying to virtual-time archival sources for tracing change longitudinally
1. Introduction
2. Trend study with radio archive sources
3. Panel study with historical corpora of written correspondence
4. Conclusion
References Chapter 7. ASPA Tools or how to measure foreign-accentedness and intelligibility in an objective manner1. Introduction
2. Existent dialectometric tools
3. English as a Lingua Franca (ELF) and the Lingua Franca Core (LFC)
4. Prototype Theory applied to (foreign) accents
5. Technical characteristics of ASPA Tools
5.1 Input/Output data
5.2 Algorithms
6. Conclusions and future perspectives
References
Chapter 8. Vowel harmony patterns in Greek dialectal child speech
1. Introduction
2. Aims and method of the present study
3. The data
4. Discussion
5. Conclusions
References

Citation preview

Language Variation – European Perspectives VII

John Benjamins Publishing Company

Studies in Language Variation

Edited by Juan A. Villena Ponsoda Francisco Díaz Montesinos Antonio Ávila Muñoz Matilde Vida Castro

22

Language Variation – European Perspectives VII

Studies in Language Variation (SILV) issn 1872-9592 The series aims to include empirical studies of linguistic variation as well as its description, explanation and interpretation in structural, social and cognitive terms. The series will cover any relevant subdiscipline: sociolinguistics, contact linguistics, dialectology, historical linguistics, anthropology/anthropological linguistics. The emphasis will be on linguistic aspects and on the interaction between linguistic and extralinguistic aspects — not on extralinguistic aspects (including language ideology, policy etc.) as such. For an overview of all books published in this series, please see benjamins.com/catalog/silv

Editors Peter Auer

Universität Freiburg

Frans Hinskens

Meertens Instituut & Radboud Universiteit Nijmegen

Paul Kerswill

University of York

Editorial Board Arto Anttila

Johannes Kabatek

Gaetano Berruto

Pieter Muysken

Jenny Cheshire

Pia Quist

Katie Drager

Anne-Catherine Simon

Katarzyna Dziubalska-Kołaczyk

Sali A. Tagliamonte

Jürg Fleischer

Øystein Alexander Vangsnes

Peter Gilles

Juan A. Villena Ponsoda

Stanford University Università di Torino University of London University of Hawai‘i at Mānoa Adam Mickiewicz University in Poznań Philipps-Universität Marburg University of Luxembourg

Universität Zürich

Radboud University Nijmegen University of Copenhagen Université catholique de Louvain University of Toronto

UiT The Arctic University of Norway Universidad de Málaga

Brian D. Joseph

The Ohio State University

Volume 22 Language Variation – European Perspectives VII Selected papers from the Ninth International Conference on Language Variation in Europe (ICLaVE 9), Malaga, June 2017 Edited by Juan-Andrés Villena-Ponsoda, Francisco Díaz-Montesinos, Antonio-Manuel Ávila-Muñoz and Matilde Vida-Castro

Language Variation – European Perspectives VII Selected papers from the Ninth International Conference on Language Variation in Europe (ICLaVE 9), Malaga, June 2017 Edited by Juan-Andrés Villena-Ponsoda Francisco Díaz-Montesinos Antonio-Manuel Ávila-Muñoz Matilde Vida-Castro University of Malaga

John Benjamins Publishing Company Amsterdam / Philadelphia

8

TM

The paper used in this publication meets the minimum requirements of the American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.

doi 10.1075/silv.22 Cataloging-in-Publication Data available from Library of Congress: lccn 2017007924 (print) / 2017032191 (e-book) isbn 978 90 272 0417 2 (Hb) isbn 978 90 272 6207 3 (e-book)

© 2019 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Company · www.benjamins.com

Table of contents

Preface Juan-Andrés Villena-Ponsoda, Francisco Díaz-Montesinos, Antonio-Manuel Ávila-Muñoz and Matilde Vida-Castro

vii

Introduction Juan-Andrés Villena-Ponsoda, Francisco Díaz-Montesinos, Antonio-Manuel Ávila-Muñoz and Matilde Vida-Castro

1

Chapter 1 Language hybridism: On the origin of interdialectal forms Manuel Almeida Chapter 2 Of clocks, clouds and sound change Frans Hinskens Chapter 3 Evaluations of foreign accent in a purist speech community: The case of Iceland Stefanie Bade Chapter 4 C’era i fascisti e i tedeschi: Instances of linguistic simplification in a corpus of Italiano popolare Silvia Ballarè and Eugenio Goria

9

27

53

71

Chapter 5 Language change caught in the act: A case study of Frisian relative pronouns 85 Jelske Dijkstra, Wilbert Heeringa, Emre Yılmaz, Henk van den Heuvel, David van Leeuwen and Hans Van de Velde Chapter 6 Virtual sociolinguistics: From real-time surveying to virtual-time archival sources for tracing change longitudinally Juan Manuel Hernández-Campoy

103

vi

Language Variation – European Perspectives VII

Chapter 7 ASPA tools or how to measure foreign-accentedness and intelligibility in an objective manner María Ángeles Jurado-Bravo and Gitte Kristiansen Chapter 8 Vowel harmony patterns in Greek dialectal child speech Ioanna Kappa and Marina Tzakosta Chapter 9 Tracking change in social meaning: The indexicality of “damped” /i/ in rural Sweden Jenny Nilsson, Therese Leinonen and Lena Wenner

119

133

145

Chapter 10 Slit-t in Dublin English Fergus O’Dwyer

159

Chapter 11 Panel and trend studies: Evidence from Brazilian Portuguese Maria da Conceição de Paiva and Maria Eugênia Lammoglia Duarte

175

Chapter 12 Quotative variation in Bernese Swiss German Christa Schneider, Sarah Grossenbacher and David Britain

191

Chapter 13 Dialect levelling or shift: Lexical outcomes of Štokavian–Čakavian contact in Dalmatia Ivana Škevin Rajko and Lucija Šimičić

203

Chapter 14 Complementing in another language: Prosody and code-switching Jonathan Steuck and Rena Torres Cacoullos

217

Chapter 15 The past perfect in Cypriot Greek: Innovation because – or irrespective – of contact? Stavroula Tsiplakou, Spyros Armostis, Spyridoula Bella, Dimitris Michelioudakis and Amalia Moser

231

Index

245

Preface Juan-Andrés Villena-Ponsoda, Francisco Díaz-Montesinos, Antonio-Manuel Ávila-Muñoz and Matilde Vida-Castro Universidad de Málaga

This book collects fifteen presentations given at the Ninth International Conference on Language Variation in Europe (ICLaVE 9), held in Malaga, Spain, June 6–9, 2017. On the one hand, the invited lectures of Manuel Almeida (Universidad de la Laguna, Spain) and Frans Hinskens (Meertens Instituut and Vrije Universiteit Amsterdam, Netherlands); and on the other, a blind-reviewed selection of papers given at the conference. The International Conference on Language Variation in Europe, ICLaVE, is one of the most important fora for research in the area of language variation in Europe. It aims at bringing together scholars of languages or language varieties spoken in Europe with the purpose of discussing empirical, methodological and theoretical issues in the study of language variation and change both on the continent and all over the world in areas where European languages have extended. Scholars in any area of linguistics, such as historical linguistics, psycholinguistics, dialectology, sociolinguistics, phonetics, formal linguistic theory, etc. are invited to submit their proposals. ICLaVE started as a biennial conference in 2001 in Barcelona (Spain) and has been held in several European countries: Sweden, Norway, Denmark, Germany (twice), Cyprus, Netherlands, and then again Spain in Malaga. A permanent international scientific committee of ten members from several European countries supports the conference, which is also held by the local organising committee:

Local committee Juan-Andrés Villena-Ponsoda. Chair (General Linguistics) Francisco Díaz-Montesinos (Spanish Linguistics) Antonio-Manuel Ávila-Muñoz (General Linguistics) Matilde Vida-Castro (General Linguistics) Gloria Guerrero-Ramos (General Linguistics) Manuel-Fernando Pérez-Lagos (General Linguistics) María-Clara von Essen. Secretary (General Linguistics).

viii Juan-Andrés Villena-Ponsoda et al.

The ICLaVE 9 local committee and the University of Malaga felt lucky to be able to host such a relevant event. (The University of Malaga has always been an open, friendly and warm space eager to harbour all sorts of scientific activities conceived as means of understanding and debate.) Analysis of similarities and differences among both European languages and speech communities all over the world allows researchers to discover a basic underlying commonality of mechanisms and priciples which is worth analysing. This was, in fact, the major aim of the group of people who attended the conference. Being, of course, a space for discussion and exchange of research data and results between specialists, this conference indirectly intended also to help ordinary people to be able to give or hear an answer to some questions that have special relevance in communities like our own, where significative social and linguistic inequalities exist. Why are speech varieties at different levels on a hierarchy of social prestige? Is the case of the Andalusian varieties so specific, in which lower-class speakers are treated as bad speakers of the standard language and not as speakers of their own local and regional variety? Why do the Andalusian middle-class speakers choose phonological, syntactic or lexical variants that make them converge towards the standard language? Many people from many countries share questions like these. Speakers need answers and conferences like ICLaVE 9 can contribute to offering some of them. Since it would probably be redundant to comment on the relevance of ICLaVE in the context of language variation and sociolinguistic research, we would rather refer to the reasons of our satisfaction to have organised the conference and be now the editors of this volume. First, this was the second time ICLaVE is held in Spain – the initial edition was at the Universitat Pompéu Frabra in Barcelona in 2001 – and it is perhaps a good opportunity to remember our late colleague and friend, Mayte Turell, who played an important role in the original organisation not only of the Barcelona conference, but of the very idea of organising an international conference on the issue of the European linguistic variation and change. Second, as some of the future readers of the present book – probably the most senior of them – will remember, the last event of the very productive ESF Network on Social Dialectology on Convergence and Divergence of Dialects in a Changing Europe – which was coordinated between 1995 and 1998 by Peter Auer and Frans Hinskens – was held in Malaga in November 1998. We organised then (with Mats Thelander and Frans Hinskens) a Summer School on the same issue that closed the network activities, where we received some of the currently most relevant European young researchers when they were still PhD students. The idea of organising an international conference on language variation in Europe started in the middle of a quite remote conversation between some of the people who attended, almost twenty

Preface ix

years later, the Malaga ICLaVE 9 in June 2017: Frans Hinskens, Mats Thelander, Paul Kerswill, Mayte Turell, Roeland van Hout and Juan Villena, all of them at the time teachers at the Malaga Summer School. After almost twenty years, the history of the IClaVE conferences all across Europe made us, as members of the organising committee, feel a bit overwhelmed but, indeed, very happy. Thanks are due to all the people and institutions who were essential to make the conference possible: the University of Malaga, for all kinds of support, the Faculty of Law, that hosted the conference, the Faculty of Philosophy and Letters and the Doctorate Programme for financial support and, of course, the ICLaVE International Scientific Committee for continuous support and advice.

Introduction Juan-Andrés Villena-Ponsoda, Francisco Díaz-Montesinos, Antonio-Manuel Ávila-Muñoz and Matilde Vida-Castro Universidad de Málaga

The Malaga IClaVE 9 brought together 355 participants from universities and research centres from all around the world. We organised the structure of the conference in three different ways of presentation: panels, paper sections and posters. Proposals for the three options were selected on the basis of blind reviews. First, a total of 13 panels were selected covering issues on many different domains in the field of language variation and change, where a total of 88 presentations were given and subjected to discussion. Second, 144 regular papers, once selected on the basis of blind-review reports, were classified in 15 specific sections. Third, a total of 16 posters were selected, dealing with a relatively large variety of European languages. And last but not least, three plenary lectures were presented. We were honoured to have among us Manuel Almeida, Universidad de la Laguna, Spain, Stella-Maris Bortoni-Ricardo, Universidade de Brasília, Brazil and Frans Hinskens, Meertens Instituut and Vrije Universiteit Amsterdam, Netherlands.

Scope of the volume The fifteen contributions in this volume, which passed a rigid process of reviewing are listed below. One the one hand, we received 26 written versions of conference presentations. After intensive reviewing both by the Conference International Scientific Committee members and a large list of anonymous reviewers from many universities and research centres from all over the world, 13 papers were definitively selected. First, out of 9 papers devoted to speech variation, 4 of them (6, 7, 8, 12) extensively focus on phonological variation, including acoustic phonetic analysis, 4 on grammar (2, 3, 10, 13) and one on lexical variation (11); second, 2 papers deal with historical sociolinguistics, namely on trend and panel studies (4, 9); then, the last two focus on foreign language issues, the first one on methods of measuring interlinguistic distance in foreign language learning (5) and the second https://doi.org/10.1075/silv.22.int © 2019 John Benjamins Publishing Company

2

Juan-Andrés Villena-Ponsoda et al.

one on attitudes towards foreign-accented speech (1). An overview of the chapters is displayed below, including (between parentheses, at the end of each summary) the varieties and languages dealt with. On the other hand, chapters by Manuel Almeida and Frans Hinskens open the book. Unfortunately, our third plenary speaker, Stella-Maris Bortoni-Ricardo did not send us her chapter in time for being part of this volume. The European languages and varieties investigated include not only large language families, such as Romance (Spanish: 4, 12, and Almeida’s chapter, Italian: 2, Portuguese: 9), and Germanic (English: 8, Swiss German: 10, Frisian: 3, Icelandic: 1, Swedish: 7), but also small language families (Greek: 6, 13) or small languages (Croatian: 11). In some other cases, a comparative approach was chosen including two or more languages (English as Lingua Franca and Non-native English: 5, English and Spanish: 4, 12). Finally, Hinskens’ theoretical chapter founds his proposals basically on Germanic – German, Dutch, Afrikaans, English –, and Romance – Latin, Spanish – languages, as well as Modern Patras Greek.

Overview of the chapters 1. As reported by Stefanie Bale, the linguistic climate in Iceland has insofar been a unique one as it has for centuries housed a relatively homogenous language, Icelandic, which has been subjected to very little (regional) variation. At the same time, institutionalised language purism and the speakers’ high regard for the Icelandic standard have always gone hand in hand. Speaker evaluations have so far largely been limited to distinctions between good and bad language concerning the native variety. With recent immigration leading to Icelandic being spoken with a foreign accent, this traditional evaluation pattern along standard/non-standard distinctions for native Icelandic might make way for a different evaluation system. This would be realised along native/non-native features, thereby influencing ideas of standardness and, possibly, changing the country’s linguistic climate. In this context, the article pays special attention to the examination of whether evaluations of non-native speakers might be, firstly, dependent on the speaker’s country of origin, and, secondly, on the listener’s background, e.g. gender and age. (Icelandic.) 2. Ballarè and Goria’s article focuses on the sociolect of Italian known as italiano popolare, which is investigated with a corpus-based methodology. This is made possible thanks to the recent ParVa compilation (Guerini 2016), one of the first oral corpora of Italian that provides information on the social background of its speakers, which allows sociolinguistic analyses. The authors investigate the phenomenon of linguistic simplification, by which italiano popolare is saliently

Introduction 3

charaterised, through a case study on agreement on the existential construction, treated as a sociolinguistic variable. Corpus data allow them to validate previous assumptions, identifying significant relations between social predictors and linguistic variables. (Standard Italian, Italiano popolare.) 3. Dijkstra and associates investigated language change in Frisian relative pronouns in a corpus of Frisian radio broadcasts (1966–2015), which was collected for training and testing a bilingual Dutch-Frisian speech recognition system. The current analyses show that until the 1980s the younger generation of speakers in these broadcasts led the rise of t-full forms of the relative pronouns, a change first mentioned in a Frisian grammar at the end of the 19th century. This change in Frisian speech towards t-full forms was also observed in the literature from the 1950s. However, the corpus data further show that from the 1980s onwards the younger generation reversed this change and increasingly started using the t-less relative pronouns, a change that was previously observed in the 1990s. So in spite of the limitations of the radio speech corpus, Dijkstra and associates were able to catch language change in the act (twice). (Frisian.) 4. In his paper, Hernández-Campoy underlines the relevance of archived radio recordings and historical corpora of written correspondence as linguistic data for both cross-sectional and longitudinal studies on language variation and change. With this aim, he presents the results of two studies carried out through the longitudinal simulation of real sociolinguistic data thanks to the use of archival sources that allow the comparison in different and distant points in time. These archival sources are shown to be crucially instrumental for the historical reconstruction of language change processes, enabling us to trace sociolinguistic patterns of language developments and thus contributing to our progress in understanding its mechanisms and motivations more accurately and immediately. (English, Murcian dialect, Castilian Standard Spanish.) 5. Jurado-Bravo and Kristiansen present ASPA Tools (Accented-Speech Phonetic Alignment), a web application which measures phonetic distances between foreign-accented speech and a standard pronunciation. Unlike existing dialectometric instruments, ASPA Tools measures the degree of intelligibility of non-native English speech in relation to English as a Lingua Franca (ELF). Since foreign accents have barely been studied using dialectometric methods, the creation of this application will contribute to the development of dialectometry by exploring the measurement of a subjective dimension, i.e. intelligibility by means of quantitative methods. Furthermore, ASPA Tools objectively measures the prototypical pronunciation of the group, which allows researchers to analyse the most salient deviations from the ELF standard. (Non-native English, English as a Lingua Franca.)

4

Juan-Andrés Villena-Ponsoda et al.

6. Kappa and Tzakosta’s study aims to test the claim that Vowel Harmony (VH) is universally motivated in child speech and facilitates phonological development in harmonic and non-harmonic languages. The analysis of a corpus of Greek L1 dialectal developmental data shows that in a non harmonic language, VH is phonologically conditioned, applies within the domain of phonological word (ω), is determined by prosodic and positional prominence effects; stress, directionality and sonority, the major principles which govern VH. Among the latter, stress and directionality are the major VH cues, while sonority/ markedness becomes effective in later stages of acquisition. The analysis supports the claim that not only does VH seem to be universal in nature, but it may also affect the order of acquisition of the Greek vocalic system in child speech. These effects are evident in the distinct developmental paths adopted by different children (inter-language) or attested in the speech of one particular child (intra-language). (Crete dialect, Standard Greek.) 7. Nilsson, Leinonen and Wenner discuss change in social meaning for a specific linguistic feature. They investigate two rural areas where the feature “damped” /i/ has been part of the traditional dialect. Through a combination of production, perception and attitude tests, they study how the social meaning of this variant has changed over time from indexing place towards indexing urban and modern. There are differences in use that reflect the complexity of the connotations of the variant. The authors show that the differences in the change in use in the two investigated speech communities reflect the attitude towards the speakers’ own dialect, and that speakers might have unconscious connotations that are undergoing change. (Rural Swedish dialects, Standard Swedish.) 8. O’Dwyer’s chapter presents a sociopragmatic as well as an acoustic analysis of a phonetic feature, the alveolar slit fricative allophone of word final /t/, used by English speakers in Dublin. Data show evidence that the slit realisation is more frequent when individuals inform the interviewer about specific subjects (and therefore they adopt a position of knowledge) in comparison to situations where talk is more social.The detailed analysis leads to a better understanding of the relationship between linguistic variants and the social side of interactions. Discourse functions clustering around the slit-t occurrences indicate that a speaker will adopt a sociopragmatic position, like emphasising a point. The author’s interpretations characterise the interactions where slit-t tokens are found as epistemically-based, inferring how speakers position themselves and others. The conclusions include a discussion of how ethnographically informed, qualitatively-skewed mixed methods can elucidate the social meaning of linguistic variants. An understanding of how phonetic variation is produced and performed should involve the integration of ethnographic approaches, to investigate the indexical meaning of particular phonetic variants. Framed in

Introduction 5

the so-called third wave, the author proposes an understanding of the salient attributes and ways of being in this specific context, and how they relate to available identities and linguistic variants. (Irish English.) 9. Paiva and Duarte discuss the power of the apparent time construct to explain some ongoing changes in Brazilian Portuguese. They analyse three variable processes in the spoken variety of Rio de Janeiro: the anterior glide deletion in diphthong [ey], the replacement of preposition A by PARA (‘to’) in dative complements and the loss of null referential pronominal subjects. By combining cross-sectional and longitudinal evidence, the authors show that the tendency observed in apparent time is confirmed for the replacement of A by PARA and the implementation of overt referential pronouns. For these two phenomena there is evidence of irreversible communal changes. In the case of [y] deletion, the predictions obtained in apparent time are only partially confirmed. The results from the longitudinal analyses suggest the importance of structural context in the direction of this process. Glide deletion proceeds in flap environments, but is reversed before palatal fricatives both in the community and in most individuals. The authors explain this reversal as a generalisation of a larger tendency in Rio de Janeiro, to insert the glide in another similar phonetic contexts. (Brazilian Portuguese.) 10. Schneider, Grossenbacher and Britain present a first variationist analysis of the quotation system of Bernese Swiss German. In order to do this, they recorded 26 working class young adults from the western part of the city of Bern – an ethnically diverse area. In doing so, they reveal the linguistic and social constraints that shape the quotative system in Bernese German. Importantly, the elements in the quotative frame are highly variable: the presence of a verb or subject, for example is not compulsory. Given the focus on the role of so in research on German quotatives, the idea was to focus on so and the social and linguistic constraints operating upon it: So is most commonly used by female speakers, and in contrast to the stereotype, mainly used by Swiss speakers, rather than speakers with a migration background. Additionally, the analysis revealed that it is often used when no verb is present, and with first and third person singular subjects. Finally, the study also highlights and questions the degree of comparability with the patterning of the innovative be like quotative in English. (Bernese Swiss German.) 11. Škevin and Šimičić address the southern Čakavian dialects spoken in Dalmatia. These relatively innovative varieties have traditionally undergone heavy Romance influence both on lexis and syntax, though locally specific Romance traits are increasingly being levelled out. Four Čakavian dialect varieties are examined in order to see if they are affected – and to what extent either by levelling (and subsequent integration into the Regional Dalmatian Koine (RD)

6

Juan-Andrés Villena-Ponsoda et al.

or shift (i.e. convergence towards Croatian Standard, CS). To do so, first, an apparent-time analysis of reported use and familiarity with Romance lexical variants was carried out to determine the degree of convergence towards both RD and CS. Second, a comparative analysis of the collected data in four originally Čakavian settings was also accomplished. Results show that (1) certain parts of the lexicon are more resistant to convergence than others, (2) there are preferred models in the processes of levelling and change, and (3) lexical stability and convergence depend on extralinguistic factors such as geographical, social and linguistic isolation (Čakavian dialects, Standard Croatian, Dalmatian regional koine). 12. Given that like monolingual speech, bilingual speech is structured not only syntactically but also prosodically, Steuck and Rena Cacoullos, using a corpus of spontaneous code-switching transcribed in Intonation Units (IU) examine the alternation between languages at the boundary of main and complement clauses. Though the alternation goes evenly in both directions (English to Spanish and Spanish to English), bilingual complementation shows a distinct prosodic pattern. In monolingual complementation, main and complement clause verbs tend to be integrated in the same IU, but when the switching of languages occurs at the clause boundary, the complement verb tends to be in a separate IU from the main verb. Interestingly, when switching between languages occurs elsewhere clause-internally but not at the clause boundary the tendency for prosodic integration is similar to monolingual speech. Therefore, the issue is not code-switching per se. Rather, bilingual complementation brings to the fore the equivalence constraint, whereby code-switching is avoided at points of word string mismatches between the languages (Poplack 1980). The mismatch here is the presence of the complementiser, which is variable in English, making the boundary between main and complement clause a point of variable equivalence. The authors conclude that bilinguals mitigate variable equivalence through prosodic distancing of code-switching boundaries. (English, Spanish.) 13. Stavroula Tsiplakou and associates address the issue of Past Perfect in Cypriot Greek. In previous research it was argued that innovative periphrastic tenses, Present and Past Perfect, are emerging in the Cypriot Greek koine. Standard Greek has periphrastic Present, Past (and Future) Perfect A, formed with the auxilary ˈexo ‘have’ and an uninflected verb form that is only marked for perfective aspect. As regards its semantics, the Standard Greek Past Perfect may mean past anterior or remote past. In Cypriot Greek, which does not have periphrastic tenses of type A, these functions are typically expressed by Simple Past. However, an innovative Past Perfect tense is emerging in the Cypriot Greek koine, which moreover seems to be fully integrated in the dialect phonologically and syntactically. The quantitative findings presented in this paper

Introduction 7

show that the innovative Cypriot Greek Past Perfect is still a Simple Past as regards its semantics, but with the additional pragmatic function of focalising an important point in the narrative. Speakers of Standard Greek who participated in this study do not accept this use of the Past Perfect in Standard Greek; however, the study revealed that there are other emergent innovative uses of the Past Perfect in Standard Greek, such as its use to round off a narrative with a verb denoting an emotive state or its use to denote recent past. The findings suggest that innovation is partly a result of contact and partly an independent development. (Cypriot Greek, Standard Greek.) Manuel Almeida (Language hybridism: On the origin of interdialectal forms) addresses interdialectalisms (a topic that has hardly been analysed from a theoretical point of view), that is, linguistic forms created from other forms or structures already existing in two dialects. An example of these hybrid forms is the variant [ɣ] of /u/, which, according to the proposal of Chambers and Trudgill (2004) [1980]: 128; also Britain 2014: 6) has arisen in East Anglia and the East Midlands as an intermediate ´compromise´ form between [ʌ], characteristic of southern dialects of England, and [ʊ], characteristic of the northern dialects. Two theories about the origin of these forms have been proposed: they can be seen as a strategy of neutrality or as the result of the imperfect learning of a second dialect. Almeida proposes a methodology that allows determining in specific cases, which interpretation is the most appropriate. To this end, he focuses on two interdialectal forms of Canarian Spanish: [cʃ], created from the vernacular form [c] and the standard variant [tʃ], and [hx], from the vernacular variant [h] and the standard [x]. Almeida analyses various aspects, such as the phonetic distance between vernacular and standard forms, linguistic uses and sociolinguistic attitudes towards the Canarian variety and the standard. He concludes that [cʃ] may have arisen as a consequence of the imperfect learning of the corresponding standard form, while [hx] may have been created in order to project a dual cultural identity, that is, to express feelings of identification with both Canarian and national culture. (Canarian dialects, Castilian Standard Spanish.) Frans Hinskens (Of clocks, clouds and sound) proposes a synthesis of sociolinguistic, formal theoretical and cognitive approaches to account for the typical diachronic development of sound change from phonetic variation to lexicalisation. The synthesis connects Usage-based Phonology and Exemplar Theory by way of sociolinguistic insights into the embedding of change in the relevant parts of the language system and the relevant sectors of the speech community. The concept of awareness is one of the pivot points connecting sociolinguistic insights into style variation with the productivity and predictability (or exceptionlessness) of a phonological rule and thus to the theory of the life cycle of sound change as developed

8

Juan-Andrés Villena-Ponsoda et al.

in Lexical Phonology and Stratal Optimality Theory. From the proposed synthesis a set of well-founded hypotheses is derived – partially obtained from a rigorous inductively-conducted revision of previous work – aiming at both ordering/testing the large number of the available results and at designing new studies. (Dutch, English, German, Spanish.) ICLaVE 9 took place in the context of the DGICYT research project on the socio­linguistic patterns of Castilian Spanish (ECOPASOS, FFI2015-68171-C5-1) and FEDER funds. We are grateful to the ICLaVE International Scientific Committee both for their advice, and for their thorough chapter revision. Several anonymous reviewers have been also involved in the long and careful process of evaluation of the chapters included in this volume, and of course of the papers submitted but eventually not accepted. Special thanks are also due to Godsuno Chela-Flores for his invaluable help.

Chapter 1

Language hybridism On the origin of interdialectal forms Manuel Almeida

Universidad de La Laguna

When two languages or varieties come into contact different processes are activated, in which several linguistic, social and cultural aspects are involved. The study of these processes is interesting because it leads to a better understanding of the mechanisms underlying language change and also of the society and culture where change occurs. In certain situations of dialect contact, new forms emerge, which did not previously exist in any of the established dialects, but have been created from elements (features, forms, structures) already existing: interdialectalisms. In this article, various theories about the origin of these forms are analysed with respect to two phonetic interdialectal forms in Canarian Spanish, both created with features from both the regional dialect and standard Spanish. Keywords: dialect contact, language change, hybridism, interdialectalism, Canarian Spanish

1. Introduction The development of hybrid linguistic forms, constructed from elements of two or more existing languages or dialects, is a process that can be observed in many speech communities.1 In the simplest cases, speakers take linguistic features, words and structures from other systems or varieties and adapt them, totally or partially, to theirs (interdialectal forms, grammatical interferences). With the passage of time, some of these forms can be so integrated into the language that receives them that virtually any trace of their origin is erased, so that their hybrid character is lost. All this suggests that the mixture of forms and linguistic structures can be seen, from a general perspective, as the result of perfectly natural processes among human groups and as a manifestation of speakers’ creativity. 1. I wish to express my thanks to Isabel Molina-Martos, Frans Hinskens, and the editors of this volume for their valuable comments to a previous version of this article. https://doi.org/10.1075/silv.22.01alm © 2019 John Benjamins Publishing Company

10

Manuel Almeida

In situations of dialectal contact, one of the most interesting manifestations of hybridism is the development of mixed forms and structures created from existing ones in two dialects: interdialectalisms.2 For Berruto (2005), the study of these forms is important because (1) they do not exist in the original dialects, being just a consequence of dialect contact, (2) imply a degree of fusion between two grammars, and (3) reveal that dialects and standard varieties can be present simultaneously in discourse. The third aspect pointed out by Berruto suggests that interdialectalisms imply the contact between a dialect and a standard variety. However, the participation of the standard to create an interdialectal form is not really necessary. Anyway, those interdialectal forms in which the standard intervenes can be seen as manifestations of an incomplete standardisation. In other dialects, as it happens with certain Andalusian varieties (Villena-Ponsoda 2000 and Villena-Ponsoda and Ávila-Muñoz 2014), hybridisation can be manifested through the development of tertiary dialects, in which characteristics of standard varieties of northern and central Spain (those closest to standard Spanish) are adapted to the pronunciation of southern dialects. This process results in the formation of a variety of the standard with abundant regionalisms and features of general Spanish: deletion of consonants at the syllable coda, loss of intervocalic /d/, etc. Both the development of interdialectal forms and tertiary dialects reveal that the change from the vernacular to the standard is not a linear and mechanical process, but a complex movement involving linguistic, social, psychosocial, cultural and (probably) geographical factors. Interdialectalisms have been analysed from two perspectives: (1) as neutral forms, compared to the variants that served as a model (Chambers and Trudgill 2004 [1980: 110–111]), and (2) as the result of imperfect learning of a second dialect (Trudgill 1986: 58–63). The aim of this research is to evaluate the explanatory capacity of those two interpretations through the analysis of two phonetic interdialectal forms of Canarian Spanish, both created from a vernacular variant and a standard Spanish one. The first hybrid form is the result of the combination of some phonetic features of the plosive consonant /c/, which tends to be voiced in the regional dialect, and the corresponding standard Castilian voiceless affricate /tʃ/ (in words like coche ‘car’, ancho ‘wide’). The second variant has arisen by contact between the fricative glottal /h/, a vernacular form, and the velar fricative /x/, a standard form (in words like gente ‘people’, naranja ‘orange’) (Table 1). For a better understanding of the nature of these two cases of phonetic hybridisation, the following aspects will be described: (1) different interpretations of hybrid forms in monolingual and multilingual communities (Section 2), (2) theoretical

2. Different examples of intermediate forms can be seen in Siegel (2010: 56–82).



Chapter 1.  Language hybridism

Table 1.  Vernacular, standard and hybrid variants of two linguistic variables  

 

Vernacular

Standard

Hybrid

(c)

coche ‘car’ ancho ‘wide’ gente ‘people’ naranja ‘orange’

[ˈkoce] [ˈanco] [ˈhente] [naˈɾanha]

[ˈkotʃe] [ˈantʃo] [ˈxente] [naˈɾanxa]

[ˈkocʃe] [ˈancʃo] [ˈhxente] [naˈɾanhxa]

(h)

and methodological aspects such as phonetic similarities between the new variants and the vernacular and standard forms, and (3) use of the different variants by a series of informants, and kinds of attitudes towards the Canarian and standard varieties (Section 3). 2. Theoretical aspects: The origin of interdialectal forms As previously stated, interdialectal forms can be explained fundamentally from two perspectives: on the one hand, as a strategy of neutrality, and, on the other hand, as a consequence of the imperfect learning of a second dialect. 2.1

The neutrality hypothesis

In an analysis of the variable (u) in East Anglia and the east of the Midlands (UK), Chambers and Trudgill (2004 [1980]: 110–111) observed that northern dialects are characterised by a predominance of the variant [ʊ] and southern dialects by a predominance of the variant [ʌ]. Between both regions there is a transition zone where two types of dialects coexist: mixed dialects, where speakers use the two variants in an alternative way, and fudged dialects, which are characterised by having developed an intermediate variant, [ɣ]. This new creation is a central not rounded vowel as [ʌ], but higher, midway between [ʌ] and [ʊ], and can be seen as a solution that allows individuals to place themselves at the same time on both poles of the phonetic continuum or conversely to avoid being identified only with one dialect. In other words, the innovative form can be seen as a neutral variant unlike the more marked forms that served as a model. 2.2

The imperfect learning hypothesis

Unlike the previous hypothesis, Trudgill (1986: 58–63) considers that interdialectal variants can be explained as a consequence of the imperfect learning of a second dialect. When speakers of a dialect try to imitate certain forms of another, some kind of failure occurs during the learning process that makes individuals reproduce

11

12

Manuel Almeida

incorrectly the forms they attempt to imitate. In this way a new linguistic form is created that doesn’t occur in either of the two dialects and yet has characteristics of both the vernacular and the target forms. In this respect, interdialectalisms would have an origin similar to that of interlanguage forms created during the process of learning a second language (Selinker 1972). For Selinker, when adults learn a second language and try to communicate in it, some forms and structures they produce do not completely coincide either with the target language or with the native, substrate one. These new forms and structures can be more or less permanently stable in their use and even become fossilised. 2.3

The interdialectal forms in Canarian Spanish

With respect to the two hybrid forms analysed in Canarian Spanish, a first approximation suggests that the interdialectal forms of (c) are most likely the consequence of imperfect learning of a second dialect. By contrast, the interdialectal forms of (h) can be viewed as an expression of neutrality, as proposed by Chambers and Trudgill (2004 [1980]: 110–111). However, the idea of neutrality that will be adopted here will be more restrictive than that defended by Chambers and Trudgill: interdialectalisms can be seen as a voluntary choice of speakers in order to project a social image that allows them to show simultaneously their identification with the Canarian culture and with the Spanish national one. For Chambers and Trudgill (2004 [1980]: 110) interdialectal forms represent a way “of being at neither pole of the continuum or conversely of being at both poles at once”. While the first part of the hypothesis of neutrality (“a way of being at neither pole of the continuum”) seems to emphasise the evasive character of the selection, the second one (“or conversely of being at both poles at once”) seems to emphasise its affirmative character. To a certain extent, these are two ways of seeing the same phenomenon; however, the interpretation that is to be defended in this article is that of dual cultural identity. The emphasis on the dual identity has also been proposed for similar linguistic solutions resulting from language contact. For instance, in a study of the Pakistani community in Glasgow, Stuart-Smith, Timmins and Alam (2011) argue that the use of hybrid forms by these speakers allows them to express both their Asian identity and their connection with their host community. This strategy is observed in an analysis of variation of syllable-initial /l/. In the Glasgow dialect, non-Asian individuals pronounce dark (velarised) /l/. However, Pakistani speakers, who do not have this sound in their mother tongue (Punjabi/Urdu), when speaking English come to employ a variant which is not as dark as that of Glasgow, but which is not as clear as that of their mother tongue, that is to say, they create a hybrid form. Muysken (1979, 1996) has proposed a similar interpretation in relation to media lengua, a

Chapter 1.  Language hybridism 13



mixed language spoken in Central Ecuador: the probable origin of this language lies in a redefinition of traditional identity by the Quechua peasant population. By incorporating Spanish into their native language, these individuals express their desire to identify with both cultures: Quechua, with a more rural orientation, and Spanish, usually associated with the urban world. When faced with this last hypothesis two aspects should be pointed out. The first one is that the interpretation supported here requires that individuals have positive attitudes towards each of the two dialects involved in the hybridising process (Myers-Scotton 1993: 12). The second one is that language hybridisation can be viewed as the first step towards a complete process of acculturation (Muysken 1979), that this, of approximation to the standard variety. The idea that the two interdialectal forms of Canarian Spanish have different origins is supported on the following aspects. According to the bibliography on Canarian Spanish, the phonetic differences between standard and vernacular forms are larger in the case of the plosive than in the case of the glottal (Section 3). Therefore, it will be easier for speakers to pronounce the standard variants of (h) than those of (c). On the other hand, if speakers can pronounce standard [x] but instead create an intermediate sound, this decision may have not to do with imperfect learning, but with a different reason: it can work as an expression of a dual cultural face. 3. Analysing interdialectal forms 3.1

Linguistic variables

The linguistic variables to be analysed are (c) and (h). Their phonetic characteristics of the variants of each variable will be described below. 3.1.1 (c) In standard Spanish [tʃ] is a prepalatal affricate consonant, with an unvoiced plosive phase that is about 25–30% longer than the fricative phase (Gili 1923; Navarro-Tomás 1918). The average duration is: 92.5 milliseconds for the close phase and 73.6 milliseconds for the fricative, the total duration of the consonant being 166.1 milliseconds (Quilis 1981: 259). In non-standard pronunciations the duration of each phase is subject to wide individual variation. The speech style can also alter the durational values of the two consonant sections, so that the emphatic pronunciation favours the lengthening of closure, while the more relaxed styles influence the lengthening of the fricative phase (Alvar 1968; Gili 1923; Navarro-Tomás 1918).

14

Manuel Almeida

The corresponding sound in the Canary Islands has the following characteristics (Almeida and Díaz-Alayón 1989: 36–37; Alvar 1959: 39–40; Catalán 1960; 1966; Lorenzo-Ramos 1976: 59–63; Quilis 1981: 267; Alvar and Quilis 1966; Trujillo 1980: 87–89): a. It is usually articulated further back than the Castilian variant, so it tends to be mid or post-palatal. b. The surface of contact between the tongue and the palate is normally larger in the Canarian articulation than in the standard consonant. c. It is usually voiced. It should be noted that in the Spanish spoken in the Canary Islands all the unvoiced plosive phonemes, /p, t, c, k/, are affected by voicing, but this is more advanced in /c/. d. The sound is perceived more as a stop than as an affricate, since the fricative phase is either deleted or is extremely short (Table 2). According to Quilis (1981: 259), in Spanish a consonant is only perceived as affricate if the fricative phase lasts at least 50 milliseconds, a requirement that is not fulfilled by any of the consonants in Table 2. Alvar and Quilis (1966), in a comparative study between the Canarian sound [c] and the Castilian [tʃ], point out that what precisely characterises the vernacular Canarian sound is the fact that there is a great disproportion between the plosive and the fricative phases, so that the first one can be twice or three times as long as the second (Figures 1 and 2). Table 2.  Duration of the plosive and fricative phases and total duration of the vernacular Canarian [c] (duration expressed in milliseconds)  

Plosive

Fricative

Total

Reference

Morro Jable (Fuerteventura) Tuineje (Fuerteventura) Fuerteventura Stressed Unstressed Santa Cruz de Tenerife Stressed Unstressed Gran Canaria Formal style Informal style

 62.0  84.0

32.0 24.0

 94.0 108.0

Alvar and Quilis (1966: 339) Alvar and Quilis (1966: 339)

129.4 130.0

22.2 23.4

151.6 153.4

Dorta (1997: 72) Dorta (1997: 72)

 66.8  63.5

22.2 27.7

 89.0  91.2

Almeida (1992: 55) Almeida (1992: 55)

110–120 70–90

30–40 30–40

140–160 Almeida (1989: 51) 100–130 Almeida (1989: 51)

Chapter 1.  Language hybridism 15



a

l

a

s

‘o

c

o

Figure 1.  Sonogram of the sequence a las ocho ‘at eight o’clock’. Example of a vernacular articulation of (c): the plosive phase is much longer than the fricative one (80.8 vs 35.8 milliseconds)

e

1

t∫

a

m

‘p

u

Figure 2.  Sonogram of the sequence el champú ‘the shampoo’ pronounced by a standard Spanish speaker. The two phases of the affricate consonant can be perfectly identified. The duration of the plosive phase is approximately 25–30% longer than the fricative one: 79.9 vs 65.1 milliseconds. Long frication bands, typical of strident sounds, are also visible

16

Manuel Almeida

Since the late 50s, it has been observed in the Canary Islands that along with the vernacular [c] another variant coexists that has been denominated palatal, strident or tense, a variant that has been documented in the speech of individuals from all social groups and areas (rural and urban) (Alvar 1959: 39–40; 1968; 1972: 124–128; Alvar and Quilis 1966). However, this variant tends to be more frequent among individuals from the highest social levels, which leads Alvar (1972: 127) to think that it probably arose due to the influence of standard Spanish. Almeida (1992) has also documented the innovative variant in Santa Cruz de Tenerife, whose main characteristic is that it is longer (o tenser) than the vernacular one. If we compare the durations of the closure displayed in Tables 2 and 3 it can be observed that they do not significantly differ in both cases: 63.5 and 66.8 milliseconds (for the unstressed and stressed variants, respectively) in the vernacular sounds (Table 2) and 54.2 and 70.4 milliseconds (unstressed and stressed variants) in the innovative ones (Table 3). As far as the duration of the fricative phase is concerned, the mean length of the vernacular variants is 27.7 and 22.2 milliseconds (Table 2), whereas that of the innovative ones is longer than 50 milliseconds (Table 3). This means that while the vernacular variant continues to be a plosive sound, the innovative one has acquired affricate status. It seems, therefore, that the speakers of Santa Cruz de Tenerife, when creating the new variant, have maintained some of the characteristics of the traditional pronunciation (voicing, duration) and have changed others (the fricative phase is longer). The new variant could be seen as an attempt to imitate the standard sound, but in this speakers have not been successful. That is why Almeida (2016) considers the innovative variant to be an interdialectal or hybrid sound (Figure 3). Table 3.  Duration of the plosive and fricative phases of the innovative [cʃ] in Santa Cruz de Tenerife   Stressed Unstressed

Plosive

Fricative

Total

70.4 54.2

64.1 52.0

134.5 106.2

An analysis of the social variation of both variants (Almeida 1992, 2016) reveals that the innovative sound is more frequent among individuals from the upper and middle-upper classes, young people, and women. The change may have started among young women of the two highest social classes and from them it would have spread to the whole community. In any case, it should be noted that tense variants are still a minority in the city (29% and 26%, respectively, in the 1992 and 2016 analyses).

Chapter 1.  Language hybridism 17



d

e

‘n

o

c∫

e

Figure 3.  Sonogram of the sequence de noche ‘at night’. Example of a hybrid variant of (c). The duration of the two phases is similar: 76.4 milliseconds for the plosive section and 79.5 milliseconds for the fricative one. Although the duration of the fricative phase is above 50 milliseconds, it is also longer than the plosive phase, unlike the standard variant

3.1.2 (h) In standard Spanish there exists an unvoiced velar fricative phoneme: /x/. According to Quilis (1981: 239), this phoneme is usually non strident, which means that its spectrum shows some transverse bands of regular intensity at different frequencies. In Canarian Spanish, as in some areas of Andalusia, the corresponding phoneme is the fricative glottal /h/, which is more frequently articulated as unvoiced when it is placed word-initial and voiced when it is intervocalic. Unvoiced variants do not usually show frication marks in a sonogram, and if present, they are usually weak. Voiced variants are characterised by a series of more or less intense narrow bands normally located at the same frequency as the formants of the vowels the consonant is in contact with and similar to the ones of approximant sounds. In this way it seems that the voiced sound does not have any particular acoustic characteristic, but it takes the characteristics of the phonetic context (that is, the preceding and following vowels). Together with these two more frequent variants there are also some more sporadic allophones, as lax and deleted, as well as intermediate forms between [x] and [h] (Almeida and Díaz-Alayón 1989: 59–63). The intermediate sounds have been recorded in rural and urban areas and in any social group, although they appear to be more frequent in the speech of young people and in the highest social groups. Almeida and Díaz-Alayón (1989) suggest that intermediate variants have probably arisen as a consequence of the prestige linked to Castilian sound [x]. However, unlike [cʃ], so far no study of the intermediate variant of the

18

Manuel Almeida

glottal phoneme has been carried out because of its low frequency in everyday speech. While waiting for more detailed analyses, it can be affirmed that the hybrid variants can move between a glottal-velar or simply velar articulation, which are perceived with an intensity of frication greater than that of the vernacular variants but lower than that of the standard (Figures 4–6).

e

n

‘h

u

n

i

o

Figure 4.  Sonogram of the sequence en junio ‘in June’. Example of vernacular (h). The only visible mark in the glottal sound is a weak transverse band at approximately the same frequency as the second and third formants of the sounds the glottal is in contact with

l

a

x

e

‘m

e

l

a

Figure 5.  Sonogram of the sequence la gemela ‘the twin (fem)’ pronounced by a standard Spanish speaker. [x] shows the characteristic transverse bands of non-strident fricatives

Chapter 1.  Language hybridism 19



l

a

‘hx

e

n

t

e

Figure 6.  Sonogram of the sequence la gente ‘(the) people’. Example of hybrid (h), with a visible band at the same frequency of the two first formants of the preceding and following vowels. Some weak transverse bands, a characteristic of non strident fricatives, are also present c01-s3-1-2-fig5

3.2

Phonetic distance between vernacular and standard forms

From the information contained in Sections 3.1.1 and 3.1.2 some conclusions about the main differences between the vernacular and standard variants of (c) and (h) can be extracted. Tables 4 and 5 show that for (c) most phonetic features differ in both dialectal varieties, while in the case of the glottal although some features are completely different (such as the one related to the place of articulation), most of them can be present in both varieties. It may be questionable to have included in the palatal an aspect such as the proportion of duration between the occlusive and fricative phases, since this characteristic can be considered to be part of the occlusive/affricate character of the variant. However, the results of the hybrid variants Table 4.  Phonetic cues of (c) variants Vernacular

Standard Spanish

Prepalatal Larger contact between the tongue and the palate Voiced/Unvoiced Stop Increment of duration of the plosive phase with respect to the fricative one: 200–500%

Mid–, postpalatal Less contact between the tongue and the palate Unvoiced Affricate

25–30%

20 Manuel Almeida

Table 5.  Phonetic cues of (h) variants Vernacular

Standard Spanish

Velar Fricative Voiced/Unvoiced Presence/absence of resonance bands Presence/absence of frication marks

Glottal Fricative Unvoiced Absence of resonance bands Presence of frication marks

(Table 3) show that the consonant can be affricate without being considered as standard. Therefore it can be concluded that there is a greater phonetic distance between the hybrid variants of (c) and the corresponding standard forms than those existing between the hybrid forms of (h) and the standard fricative. 3.3

Individual variation

The kind of changes described here are more frequent in certain public or even formal contexts (such as academic debates, speeches in public institutions, radio and television programmes, etc.), and in certain social groups (especially, individuals from the middle and upper-middle classes). Both characteristics are typical of changes from above, as described by Labov (1982 [1966]: 221–226; 1991 [1972]: 178–181). With respect to the contexts of use of the two variables under analysis, the hybrid variants of (c) can be present in all kinds of situations, but the hybrid and standard variants of (h) rarely appear in informal conversations.3 That means that if we want to study (h) variation we need to focus on formal situations. A quick way of gathering speech samples from formal contexts is the recording of radio and television programmes. In this case only radio talk shows and interviews have been analysed. The recordings were made in the 2006–2010 period, and only some of the radio stations broadcasting regionally were selected: La Ser, Cadena Cope, Radio Las Palmas, and Radio Nacional de España. The sample consists of about twenty hours of recordings. Eight informants were selected for the individual analysis: 6 presenters (3 men and 3 women), 1 politician (a woman) and 1 teacher (a woman). From each informant, a maximum of 50 instances were selected. 3. A situation similar to that of (h) occurs in other phonological variables. The Canarian linguistic norm is characterised, in contrast to the standard Castilian, by two important aspects: (1) the aspirated and deleted pronunciation of syllable-final /s/ and (2) the non-distinction of /s/ and /θ/, since both are pronounced as [s] (seseo). In that way, standard pronunciations like [las ‘kasas] ‘the houses’ and [ˈθena] ‘dinner’ are normally heard in Canarian Spanish as [lah ˈkasa(h)] and [ˈsena] respectively. However, in public speeches the vernacular forms can alternate with intermediate forms like [lah ‘kasas]–[las–ˈkasah] or [ˈsθena]. These two kinds of interdialectalisms have not been analysed here because they show more complex theoretical aspects.

Chapter 1.  Language hybridism 21



The results of the analysis of individual variation are contained in Tables 6 and 7. The first interesting result is that for each variable most informants mainly use the vernacular forms, a sign of their great vitality. There is just one exception, the teacher, who uses vernacular, hybrid and standard forms of (h) in a similar way. This situation of dialectal variation contrasts with the time of Franco’s dictatorship, when broadcasting on the radio in the Canary Islands, especially in radio stations located in urban centres, presenters were forced to use the standard Castilian forms. This practice disappeared from general use with the arrival of democracy, although some professionals of radio and television continued speaking the standard variety (Yanes 2013). The percentages of the variants of each variable also reveal that standard variants of (c) are almost absent (only four cases were found). In contrast, in the case of (h) just the opposite happens: only one informant (the presenter 6) has not pronounced any standard form. The information contained in Tables 6 and 7 seems to indicate that there is a kind of implication in the relationship between the three types of variants, so that before beginning to use the standard forms, individuals usually pass through the phase of hybridity: vernacular form > hybridism > standardisation. There is just one exception: the presenter 5. It is probable that individuals do not want to make the change from the vernacular to the standard abruptly, but transitionally, so that they would use the hybrid forms first and, once they have accepted the possible implications of their decision (for example, when they evaluate the degree of acceptance and rejection of this decision), they would decide to use the standard variant. The results are in line with those described above concerning the phonetic distance of the hybrid forms: it is not difficult for Canarian speakers to pronounce [x] instead of the vernacular form [h], but it is hard for them to pronounce standard variants of the plosive palatal consonant. Table 6.  Individual variation of variants of (c)  

Vernacular

Hybrid

 

 

N

%

N

%

Total

Presenter 1 (w) Presenter 2 (w) Presenter 3 (w) Presenter 4 (m) Presenter 5 (m) Presenter 6 (m) Politician (w) Teacher (w)

33 49 34 34 50 49 23 36

 82.5  98.0  68.0  94.4 100.0  98.0  59.0  92.3

 7  1 16  2  0  1 16  3

17.5  2.0 32.0  5.6  0.0  2.0 41.0  7.7

40 50 50 36 50 50 39 39

w = woman, m = man

22

Manuel Almeida

Table 7.  Individual variation of variants of (h)  

Vernacular

Hybrid

Standard

 

 

N

%

N

%

N

%

Total

Presenter 1 (w) Presenter 2 (w) Presenter 3 (w) Presenter 4 (m) Presenter 5 (m) Presenter 6 (m) Politician (w) Teacher (w)

23 42 32 34 46 46 45 11

46.0 84.0 64.0 68.0 92.0 92.0 90.0 31.4

18  4 16 12  0  4  4 12

36.0  8.0 32.0 24.0  0.0  8.0  8.0 34.3

 9  4  2  4  4  0  1 12

18.0  8.0  4.0  8.0  8.0  0.0  2.0 34.3

50 50 50 50 50 50 50 35

w = woman, m = man

In summary, each of the two interdialectal forms maintains a different type of relationship with the vernacular and with the standard forms from which they come. The phonetic characteristics of both variants and the analysis of their uses reveal that the intermediate forms of (c) may be due to an imperfect learning of the standard pronunciation, while intermediate forms of (h) may be a consequence of a voluntary decision of speakers to project a social image that expresses feelings of identification with both regional and national cultural values. It is possible that the decision of creating a new form has had to do with the most marked character of the variants that served as a model, as pointed out by Chambers and Trudgill (2004 [1980]: 110–111), since the new interdialectal variants cannot be associated to a specific dialect yet. 3.4

The role of attitudes

As previously mentioned, the interpretation of hybrid forms as a way of expressing a dual cultural identity is only possible when individuals maintain positive attitudes towards the two varieties involved in its creation. No such study has been carried out in the present analysis, but, if it were the case, it would be very likely for the informants of this study to reproduce the results obtained in a research carried out by Morgenthaler-García (2008: 371–372) on attitudes of the Canarians towards their own dialect and towards standard Spanish. In Morgenthaler’s research it was demonstrated that when both dialects are evaluated, the Canarians maintain positive attitudes towards their own modality regarding interpersonal attributes: it is viewed as sweet, soft, close and non-aggressive, while standard was valued as insulting, aggressive and distant. With respect to social prestige, the Canarian dialect was negatively evaluated, for it was described as incorrect and rough; on the contrary,



Chapter 1.  Language hybridism 23

standard achieved a positive evaluation, and thus it was valued as correct, fine and adequate. It can be said that the values of the Canarian dialect are associated with personal attractiveness, whereas those of standard Spanish are associated with sociocultural status. Similar attitudes were described in the 1970s with regard to the relationship between Standard English and British dialects (Giles and Powesland 1975). In Labovian terms (Labov 1982 [1966]: 286–304), the Canarian variety has acquired covert prestige and the standard variety overt prestige. 4. Conclusions As a consequence of dialect contact, different outcomes have been described in European dialects, interdialectalisms being one of the most creative. Two main interpretations about the origin of these forms have been proposed up to now: functional (they serve as a strategy of neutrality) and non-functional (they are a consequence of the imperfect learning of a second dialect). However, the studies carried out on hybrid language forms have not developed a methodology that helps to decide which kind of interpretation (neutrality, imperfect learning) is the most appropriate in a specific case. In this article several strategies are proposed at this respect: (1) analysis of the phonetic distance between the two variants involved in the hybridising process, (2) types of variants used by individuals and percentages of use of each one, and (3) description of the attitudes towards the two varieties that participate in the creation of the new form. It is also proposed that the thesis that refers to the neutral character of interdialectalisms be substituted by this other one: interdialectal forms can be viewed as a way of expressing feelings of identification with the cultures associated with the two dialects involved in the process of hybridisation. The analysis carried out so far has demonstrated that not all the hybrid forms arising in a community are necessarily related to the same causes. In the Canary Islands, the hybrid forms of the palatal plosive can be explained as a consequence of imperfect learning, whereas the hybrid forms of the glottal consonant can be viewed as a response to the desire of speakers to transmit a social image where feelings of identification with both the Canarian vernacular culture and the one represented by the standard variety are manifested. The main argument to justify this interpretation is that the phonetic distance between the vernacular and the standard variants is greater in the plosive than in the glottal variable. Consequently most speakers use standard forms of (h), but hardly use standard variants of (c). This interpretation is also supported by the description of attitudes carried out by Morgenthaler-García (2008). Canarian people associate their own variety with certain attributes that can

24

Manuel Almeida

be considered a manifestation of covert prestige, while associating the Castilian variety with certain attributes usually related to overt prestige. The present analysis, while answering some questions, leads to new ones. For example, can interdialectal forms be treated in the same way as interlanguage forms, as Trudgill (1986: 62) suggests? If so, we do need to accept the existence of a third grammar, different to that of the vernacular and the standard (Tarone 1994). The consequences of such an implication would need to be considered. Another aspect to be addressed is the interpretation of hybrid forms as ways of expressing a dual identity and therefore a critique to the concept of identity related to rationalist theories (such as correlational or structural Sociolinguistics) and a support of postmodern theories, as stated by García-Calcini (2013 [2001]: 17–20). As is well known, within the rationalist framework identity is conceived as a set of fixed traits that derive from the existence of solid and perfectly constituted social categories (such as class, ethnicity, gender, generational group). From the postmodern perspective, while rejecting the existence of objective sociocultural categories (class, gender, etc.) with the argument that in today’s societies there is a greater fluidity among social groups than in the past, it is also assumed that identities are no longer fixed, but changeable and negotiable. Unlike individuals of traditional societies, embedded in a rigid system of social organisation that made perfect recognition of groups possible, in modern, extremely changing societies, individuals interact in different scenarios and can develop different identities in each one (Hall 1992). A third aspect that should be studied is the linguistic variation, since the analysis of the contexts where the interdialectal forms take place would allow us to determine how these forms are embedded in the linguistic structure. Almeida (2016) observed that the interdialectal forms of the plosive did not seem to follow a regular pattern of variation, possibly due to their hybrid character. A fourth aspect that should be analysed is what kind of relationship interdialectal forms maintain with other mixed linguistic forms, such as code-switching, or with other kinds of language creations, such as blending (motel < motor + hotel, brunch < breakfast + lunch). Finally, in order to better understand the functioning of hybrid forms, a greater variability of social groups should be taken into account. For example, it would be necessary to determine how ordinary people speak when they phone radio stations for the most diverse reasons, or when they are interviewed in the streets, at their workplaces, etc. The study of interdialectal forms has not yet gone beyond a descriptive phase, but deeper analyses that develop in the future can provide interesting theoretical insights for disciplines such as Dialectology, Sociolinguistics or Psycho­linguistics.



Chapter 1.  Language hybridism 25

References Almeida, Manuel. 1989. El habla rural en Gran Canaria [The rural speech in Gran Canaria]. La Laguna: Universidad de La Laguna. Almeida, Manuel. 1992. Mecanismos sociolingüísticos del cambio fonético [Sociolinguistic mechanisms of phonetic change]. In José A. Bartol-Hernández, Juan F. García-Santos, and Javier de Santiago-Guervós (eds.), Estudios filológicos en homenaje a Eugenio de Bustos Tovar, 51–60. Salamanca: Universidad de Salamanca. Almeida, Manuel. 2016. Constraints on language change in the Canary Islands. Sociolinguistics Symposium 21. Murcia: University of Murcia (15th–18th June). Almeida, Manuel and Carmen Díaz-Alayón. 1989. El español de Canarias [The Spanish spoken in the Canary Islands]. Santa Cruz de Tenerife: Self-publishing. Alvar, Manuel. 1959. El español hablado en Tenerife [The Spanish spoken in Tenerife]. Madrid: Consejo Superior de Investigaciones Científicas. Alvar, Manuel. 1968. Datos acústicos y geográficos sobre la ch adherente [Acoustic and geographical data on adherent ch]. In Manuel Alvar, Estudios canarios I, 71–78. Las Palmas de Gran Canaria: Excelentísimo Cabildo Insular de Gran Canaria. Alvar, Manuel. 1972. Niveles socio-culturales en el habla de Las Palmas de Gran Canaria [Socio­ cultural levels in the speech of Las Palmas de Gran Canaria]. Las Palmas de Gran Canaria: Excelentísimo Cabildo Insular de Gran Canaria. Alvar, Manuel and Antonio Quilis. 1966. Datos acústicos y geográficos sobre la “ch” adherente de Canarias [Acoustic and geographic data on adherent “ch” in the Canary Islands]. Anuario de Estudios Atlánticos 12. 337–343. Berruto, Gaetano. 2005. Dialect/standard convergence, mixing, and models of language contact: The case of Italy. In Peter Auer, Frans Hinskens and Paul Kerswill (eds.) Dialect change: Convergence and divergence in European languages, 81–95. Cambridge: Cambridge Univer­ sity Press. Catalán, Diego. 1960. El español canario entre Europa y América [Canarian Spanish: Between Europe and America]. Boletim de Filologia XIX. 319–337. Catalán, Diego. 1966. El español en Tenerife. Problemas metodológicos [The Spanish spoken in Tenerife: Methodological problems]. Zeitschrift für Romanische Philologie LXXXII. 467–506. Chambers, J. K. and Peter Trudgill. 2004 [1980]. Dialectology. Cambridge: Cambridge University Press. Dorta, Josefa. 1997. Datos acústicos y percepción de la [ĉ] adherente de Canarias y de la pre-palatal castellana [Acoustic data and perception of the Canarian adherent [ĉ] and the Castilian prepalatal consonant]. In Manuel Almeida and Josefa Dorta (eds.) Contribuciones al estudio de la lingüística hispánica. Homenaje al profesor Ramón Trujillo, 57–72. Barcelona: Montesinos. García-Canclini, Néstor. 2013[2001]. Culturas híbridas. Estrategias para entrar y salir de la modernidad [Hybrid cultures: Strategies for entering and leaving modernity]. Barcelona-Buenos Aires-México: Paidós. Giles, Howard and Peter F. Powesland. 1975. Speech style and social evaluation. London: Academic Press. Gili, Samuel. 1923. Observaciones sobre la ĉ [Observations on ĉ]. Revista de Filología Española X. 179–182. Hall, Stuart. 1992. The question of cultural identity. In Stuart Hall, David Held and Tony McGrew (eds.) Modernity and its futures, 273–316. Cambridge: Polity Press.

26 Manuel Almeida

Labov, William. 1982 [1966]. The social stratification of English in New York City. Washington, DC: Center for Applied Linguistics. Labov, William. 1991 [1972]. Sociolinguistic patterns. Philadelphia: University of Pennsylvania. Lorenzo-Ramos, Antonio. 1976. El habla de Los Silos [The speech of Los Silos]. Santa Cruz de Tenerife: Caja General de Ahorros de Santa Cruz de Tenerife. Morgenthaler-García, Laura. 2008. Identidad y pluricentrismo lingüístico. Hablantes canarios frente a la estandarización [Identity and pluricentrism: Canarian speakers facing standardisation]. Frankfurt am Main: Vervuert-Madrid: Iberoamericana. Muysken, Pieter. 1979. La mezcla de quechua y castellano. El caso de la “media lengua” en el Ecuador [Quechua and Spanish mixing: The case of “media lengua” in Ecuador]. Lexis III (1). 41–56. Muysken, Pieter. 1996. “Media lengua.” In Sarah G. Thomason (ed.) Contact languages: A wider perspective, 365–426. Amsterdam-Philadelphia: John Benjamins Publishing. Myers-Scotton, Carol. 1993. Duelling languages: Grammatical structure in codeswitching. Oxford: Clarendon Press. Navarro-Tomás, Tomás. 1918. Diferencias de duración de las consonantes españolas [Differences of duration of Spanish consonants]. Revista de Filología Española V. 367–393. Quilis, Antonio. 1981. Fonética acústica de la lengua española [Acoustic phonetics of the Spanish language]. Madrid: Gredos. Selinker, Larry. 1972. Interlanguage. International Review of Applied Sociolinguistics X (3). 209–231. Siegel, Jeff. 2010. Second dialect acquisition. Cambridge: Cambridge University Press. https://doi.org/10.1017/CBO9780511777820 Stuart-Smith, Jane, Claire Timmins and Farhana Alam. 2011. Hybridity and ethnic accents: A sociophonetic analysis of ‘Glaswegian’. In Frans Gregersen, Jeffrey K. Parrott and Pia Quist (eds.), Language variation: European perspectives III, 43–57. Amsterdam-Philadelphia: John Benjamins Publishing.  https://doi.org/10.1075/silv.7.04stu Tarone, E. 1994. Interlanguage. In R. E. Asher and J. M. Y. Simpson (eds.), The encyclopedia of language and linguistics 4, 1715–1719. Oxford: Pergamon. Trudgill, Peter. 1986. Dialects in contact. Oxford: Blackwell. Trujillo, Ramón. 1980. Lengua y cultura en Masca [Language and culture in Masca]. Santa Cruz de Tenerife: Interinsular Canaria. Villena Ponsoda, Juan A. 2000. Identidad y variación lingüística: prestigio nacional y lealtad vernacular en el español hablado en Andalucía [Identity and linguistic variation: National prestige and vernacular loyalty in the Spanish spoken in Andalusia]. In Georg Bossong and Francisco Báez de Aguilar-González (eds.), Identidades lingüísticas en la España autonómica, 107–150. Frankfurt-am-Main: Vervuert-Madrid: Iberoamericana. Villena Ponsoda, Juan A. and Antonio M. Ávila Muñoz. 2014. Dialect stability and divergence in southern Spain: Social and personal motivations. In Kurt Braunmüller, Stefen Höder and Karoline Kühl (eds.), Stability and divergence in language contact: Factors and mechanisms, 207–238. Amsterdam-Philadelphia: John Benjamins Publishing. Yanes, Julio. 2013. La locución radiofónica en Canarias durante el franquismo [The radio locution in the Canary Islands during the Franco’s regime]. Revista Internacional de Historia de la Comunicación 1 (1). 155–175.

Chapter 2

Of clocks, clouds and sound change Frans Hinskens

Meertens Instituut and Radboud Universiteit Nijmegen

The study of sound change has evolved from a heuristic tool for 19th century comparative historical reconstruction into the backbone of the rigid approach to language change developed by the Neogrammarians. In the course of the 20th and early 21st century it has become the main meeting point for a range of subdisciplines of linguistics (historical linguistics, dialectology, sociolinguistics, phonology, phonetics and cognitivist approaches to phonetic variation). This contribution sketches some of the main aspects of the approaches to sound change taken in the various corners of the field. By way of a synthesis a theory will be outlined in which three approaches to sound change dovetail to account for the huge and seemingly chaotic body of insights into the phenomenon. Empirical studies of instances of both historical and ongoing sound change in specific varieties of Dutch will serve to illustrate parts of the theory. Keywords: awareness, Exemplar Theory, generative phonology, hyperdialectism, indicator, marker, stereotype, lexically diffuse sound change, lexical frequency effects, life cycle of sound change, (mis-)perception, Neogrammarians, opacity, Optimality Theory, style, Usage based phonology, variation

1. Introduction The study of sound change (henceforth: SC) is centrestage in a range of subdisciplines of linguistics and the range keeps broadening.1 There are two main aspects to be distinguished: the process and the result, i.e. the difference between before and after. On the process-side, analyses concern the origin, implementation and spread of SC, in connection with the result, the focus is on the outcome and its 1. This contribution overlaps with Hinskens (2020). In most respects the present paper is more compact and goes less deep, but at several points it goes considerably further than the 2019 one. Many thanks to the editors, Antonio Ávila-Muñoz, Francisco Díaz-Montesinos, Matilde Vida-Castro and Juan Villena-Ponsoda, as well as to Manolo Almeida for their invaluable comments and suggestions. Of course, they are not to responsible for any remaining shortcomings. https://doi.org/10.1075/silv.22.02hin © 2019 John Benjamins Publishing Company

28

Frans Hinskens

various properties. In both dimensions, three of the main questions are (cf. Scheutz 2005: 1704): – what: does the change concern a specific sound or a natural class as in, e.g., the famous Second or High German obstruent shift? – how: did it develop internally (through innere Kausalität, as Moulton 1961 labeled it) or through contact? – why: did it result from structural forces, co-articulation or prestige-driven borrowing? These considerations and the relevant sub-disciplines can be schematically summarised as follows: Table 1.  Five sub-disciplines and their division of labour in the study of SC  

what

how

why

process

sociolinguistics

phonetics, sociolinguistics

phonetics, sociolinguistics

outcome

historical linguistics, dialectology

dialectology

phonology

The time axis is what ties together historical linguistics, dialectology and sociolinguistics: whereas historical linguistics highlights diachronic aspects (extrapolated through juxtaposition of two more synchronic states of the language system at issue), dialectology focuses on the synchronic reflection of diachrony in geography and sociolinguistics is concerned with synchronic variation. Where dialectology (often implicitly) treats dialects as homogenous systems, sociolinguistics approaches intra-dialect variation as structured heterogeneity. Phonetics zooms in at the listener (and the role of (mis-)perception; cf. Ohala 2003) and the speaker; the attention paid to articulatory gestures is growing both theoretically (Browman and Goldstein 1992) and technically (the registration of tongue movements with ultrasound techniques). Whereas phonetics deals with the hardware (physics, anatomy) and looks at speech as a phenomenon which is gradient and continuous in most respects, phonology deals with the software (psychology, cognition) and looks at the organisation of the sound structure as a system of categorical phenomena. 2. The Neogrammarian legacy Nineteenth century linguistics was largely geared towards historical reconstruction and Schleicher (1861) already discussed ‘exceptionlessly applying laws’ (my translation, FH). In the thinking of at least some Neogrammarians, the concept evolved further. The essence of Neogrammarian thinking is laid down in the following oft-cited quote from what has been referred to as the Neogrammarian manifesto:

Chapter 2.  Of clocks, clouds and sound change 29



Every sound change, insofar as it operates mechanically, proceeds according to exceptionless laws, i.e. the direction of the sound shift is always the same for everybody belonging to a speech community, except where a dialect split occurs, and all words in which the sound affected by the sound shift appears in the same context are affected by the change without exception. (Osthoff and Brugmann 1878: xii)2

In the Neogrammarian model, the effects of the sound laws or rules can be obscured by competing SC (Wang 1969), including bleeding or counter-feeding rule orderings, analogy or borrowing. An example of a competing SC concerns Medieval and Early Modern Dutch /u:/ > /y:/ > /œy/; the second step in this shift resulted in the rounded front diphthong (of which the first element is prominent). This latter change did not take place before /r/ because of the centralising effect of tautosyllabic /r/ on preceding non-low tense vowels, as in e.g. zuur ‘sour’, duur ‘expensive’, muur ‘wall’; here, centralisation (yielding VVə) bleeds diphthongisation. The following example of an instance of analogy builds on Pre-Latin intervocalic /s/ > [z] > /r/, which applied in e.g. hono:sis > hono:zis > hono:ris , ‘honor’ (sg gen), flo:sis > flo:zis > flo:ris, ‘flower’ (sg gen). Later this SC, which was exceptionless and phonetically gradient, also applied word-finally, i.e. in stems, as in honos > honor, but not in flos > *flor, hence not exceptionlessly. Finally, an example of borrowing concerns the fact that Latin rosa, ‘rose’, basis, ‘pedestal’ (from Greek), did not undergo rhotacism (/s/ > [z] > /r/), as they were borrowed after the SC had ceased to be productive, i.e. when the speakers had ceased to apply the SC to new words. These and numerous other, similar observations have led to the idée reçue that in reality exceptionless SC hardly exists. In fact, some instances of SC look rather fuzzy. In 1966 Karl Popper published an essay entitled Of clocks and clouds; in this text Popper contrasts neat, orderly and predictable phenomena with messy, chaotic and unpredictable phenomena. Many data from dialectology have made plain that SC may not be as clock-like as the early Neogrammarians conceived, but is it really an epistemic mess? To answer this (admittedly: rhetorical) question, SC will here be looked at from three different perspectives, viz. formal phonological theory, the sociolinguistic study of language variation and change, and cognitive approaches, respectively. For none of these perspectives, anything new will be presented. The innovatory aspect lies in the fact that and the way in which these perspectives will be brought together. Interlocking the three views will result in an integrated and testable theory of the diachronic development of SC and the process of its embedding in both the language system and the speech community. The cloud will not be transformed into a clock, but the mutual synchronisation of the three perspectives may bring the researcher closer to a better insight into the complexity of SC.

2. Translation in Garrett (2009: 2).

30

Frans Hinskens

3. Formal theory: Generative phonology and Optimality Theory In early linear theory (introduced in and developed out of Chomsky and Halle 1968), the locus of variation in so far as variation was at all an issue was conceived to reside in the rule machinery. Dialect variation was accounted for through rule ordering: two rules which are differently ordered constitute either two different (categorical) grammars or inherent, quantitative variation. An early example concerned Umlaut3 and lowering of back vowels in the German dialects of neighboring Schaffhausen and Kesswill in North-Eastern Switzerland (Kiparsky 1968: 178–179). An issue which is still around and which keeps exciting phonologists is opacity. Opacity comes in two types. The first is residue, i.e. forms which fit the structural description of a given rule, yet do not undergo / have not undergone the rule. The second type concerns forms which do not fit the structural description of a given rule, yet do undergo / have undergone the rule. The first type is clearly related to the concept of exceptionlessness. In formal theory, phonology is conceived as an autonomous module of the grammar. In this view, phonology is itself, in turn, organized modularly, with interfaces with the lexicon, morphology and syntax. As far as the lexicon is concerned, the segment inventory and morpheme structure constraints are relevant issues. With regard to morphology, the distinction between cyclic and postcyclic phonological rules is relevant, and on the level of the syntax, the distinction between post-lexical (or late) processes and phonetic implementation applies. 3.1

The life cycle of sound change

A very promising insight that has developed in lexical phonology and related Stratal Optimality Theory is often referred to as the “life cycle of sound change” (Kiparsky 1995; Bermúdez-Otero 2015; Ramsammy 2015). In this conception of the diachronic development of SC, the cycle is ignited by more or less random phonetic variation. Random phonetic variation can become systematic and from that point onwards it will be exceptionless. An instance of systematic, directional phonetic variation can change into a postlexical process; the main property setting apart post-lexical variation from phonetic processes is the fact that the application of postlexical processes tends to be confined to specific prosodic domains, such as the syllable onset or the end of the prosodic word. Postlexical processes can narrow to become lexical rules, rules which apply to morphologically derived words, such as the original version of Umlaut. Another example is e-prothesis in Romance 3. The fronting of a back vowel in a variety of morphologically derived environments, such as noun pluralisation and diminutive formation.

Chapter 2.  Of clocks, clouds and sound change 31



ISA TIO

N

AT IO LIS

LO G

ICA

MO RP

HO

LEX

N

TIO ISA

OG OL

ON

Epiphenomenal, gradient phonetic effect

Lexically stored information

PH

N

(as in Spanish escuela, French école, ‘school’), which originally had “phrase-level conditioning”, but the “purely syllable-structural basis” was weakened “by making it crucially dependent on word boundaries” (Janda 2003: 420). The last step is for processes or rules to be lexicalised, either in free morphemes (items) or in bound morphemes (affixes) or both. An example is again Umlaut, with or without concatenation. In English it has become marginal (goose–geese, foot–feet, but crook–*creek); in modern standard German, where it has long been productive, Umlaut is no longer synchronically predictable, cf. Fuss–Füsse, ‘foot–feet’, Grund–Gründe, ‘reason(s)’, but Hund–Hunde, ‘dog(s)’. Going from systematic phonetic variation to lexicalisation, structural predictability decreases. Figure 1, from Ramsammy (2015), visualises these and closely related insights as well as the cyclical nature of the mechanism. The development of gradient phonetic processes into systematic and directional, cognitively controlled processes is referred to as phonologisation. Once these processes stabilise, they become postlexical processes and they can lose their gradient and quantitatively variable nature, becoming categorical, as in the case of allophony. In this phase, rules can narrow down their domain of application to words (honos, but honoris) or even stems (honos > honor), hence becoming lexical (or minor) rules. Eventually they lexicalise or morphologise, i.e. they freeze and become stored in the lexical form of specific items or morphemes – and morpheme structure constraints may result.

The life cycle of phonological processes

Stem-level categorical phonological rule

AI M DO OW G

IN II DOMAIN NARROWING I

Figure 1.  The life cycle of phonological processes. From Ramsammy (2015)

BIL

RR

NA

Phrase-level categorical phonological rule

STA

N

Word-level categorical phonological rule

ISA TIO

N

Gradient phonetic process, under cognitive control

32

Frans Hinskens

At the beginning, early in the cycle, during its phonologisation and stabilisation, a process can be generalised, typically through the extension of its context of application. An example is the non-standard German fricativisation of /ɡ/ into [j] (Scheutz 2005: 1707). The process originated before palatal vowels, i.e. in words such as (1) a.

/ɡ/iessen ~ [j]iessen /ɡ/eben ~ [j]eben

‘to pour’ ‘to give’

as a process of co-articulation and assimilation. Later this process was generalised to include also contexts before non-palatal vowels, e.g.  

b. /ɡ/abel ~ /ɡ/old ~

[j]abel [j]old

‘fork’ ‘gold’

And later even to include liquids as in, e.g.  

c.

/ɡ/raben ~ [j]raben ‘to dig’ /ɡ/lauben ~ [j]lauben ‘to believe’

At the end of the cycle, i.e. after a rule has been lexicalised, it can either become extinct or it can spread in a lexically diffuse fashion. As to the first scenario, rule death (also referred to as rule loss) can trigger analogy. An example is pre Latin rhotacism /s/ > [z] > /r/, sketched above. Another example is from the history of Dutch and it concerns the historical rule of open syllable vowel lengthening (OSL) and the way its effects were undone, later. In Early Modern Dutch, OSL occurred e.g. as a result of pluralisation through the suffixation of -ə suffix (orthographically ), the final consonant was resyllabified to become the onset of the new ə-syllable. Consequently, the stem vowel was in open syllable position and was lengthened, as in (2) a.

sch[ɔ]t-sch[o ː]ten ‘shot(s)’, b[ɑ]d-b[a ː]den ‘bath(s)’ bissch[ɔ]p-bissch[oː]pen ‘bishop(s)’

but certain plurals, such as  

b. bissch[o]pen → bissch[ɔ]pen ‘bishops’

have been analogically restructured after the model of short vowel nouns such as  

c.

m[ɔ]t-m[ɔ]tten ‘moth(s)’ r[ɑ]t-r[ɑ]tten ‘rat(s)’

which had never undergone OSL. Analogical restructuring can thus undo the effects of the lexicalisation of a SC which had fallen in disuse.

Chapter 2.  Of clocks, clouds and sound change 33



In the second scenario, the SC comes back to life, but as a phonological zombie, as it were, i.e. as an undead rule, by generalising in a lexically diffuse fashion. An example concerns the two-step historical vowel shift discussed in Section 2 above, the palatalization and subsequent diphthongisation of the West Germanic long high back vowel /u:/ > /y:/ > /œy/. There are ample exceptions for each of the consecutive steps. There is an unshifted /u:/ in items such as (3) a.

smoel sm[uː]l ‘conk, physgog’; boer b[uːə]r ‘farmer’

The second step has also left residue behind, such as  

b. beduusd bed[yː]sd gruzelementen gr[yː]zelementen ruzie r[yː]zie

‘bewildered’ ‘smithereens’ ‘quarrel’

all of which have the front rounded vowel resulting from the palatalisation of /u:/. This evidence shows how both telescoped SC’s have been spreading word-by-word. The mechanism has come to a complete stop, hence a recent loan such as ‘blues’ never surfaces as *[blœys]. So after a rule has ceased to be productive, its lexicalised left-behinds can either undergo analogy (and thus disappear) or, on the contrary, start infecting unshifted items with the former SC, which then enters an afterlife as a zombie rule, i.e. as a lexically diffuse rule (cf. Kiparsky 1995). There is a third option: the mechanism starts anew. Sometimes an old, lexicalized SC becomes fully productive again. This is what is happening today with the historical vocalisation of postvocalic /l/ after low back vowels, which resulted in the diphtong /ɑu/, where German has retained the back vowel and the liquid. Examples include modern standard Dutch (4) a.

oud [ɑut] < ald ‘old’ woud [ʋɑut] < wald ‘wood’

In many varieties of spoken modern Dutch it has become productive again, also following front vowels, as in (4b).4  

b. Niels 4 Nie[w]s ‘Nils’ meel mee[w] ‘flour’

4. When she was 5 years old, the daughter of the present author used to refer to Pippi Long­ stocking’s little monkey, Mr Nilsson, as Ni[uw]sson.

34

Frans Hinskens

  5

c.

school schoo[w] ‘school’ ‘tosses about’; ‘makes a fool of ’ solt 5 so[w]t halt h[ɑw]t ‘halt; stop’

with a vocalised liquid, these variants are nearly homophonous with6    

d. nieuws meeuw e. schouw zout 6 hout

‘news’ ‘seagull’ ‘chimney’ ‘salt’ ‘wood’

respectively. So postvocalic /l/ vocalisation, which took place in Middle Dutch, was there and is back again; it has gone full circle. As the case of postvocalic /l/ vocalisation clearly illustrates, the life cycle of SC is a continuum, but one in which the two extremes (Neogrammarian SC, targeting phonemes, and lexically diffuse SC, targeting lexical items) sometimes meet, hence a cycle; there is no need for a Neogrammarian controversy (Labov 1981). The life cycle of SC is uni-directional, as the change invariably proceeds from sound structure, via grammar, to lexicon. The idea is highly comparable to Janda and Joseph’s (2003) Big Bang theory, “saying that sound changes start as phonetically determined ‘events’ in very ‘small’ environments the big bang and in the aftermath of a phonetic event there can be generalization along various lines”, viz. phonetic trajectory, phonology, morphology, social space and the lexicon (Joseph 2012: 420–421). The life cycle of SC is a common, yet complex process involving physical, cognitive as well as social mechanisms – in short, all dimensions of human nature. Because of its complexity, it proceeds slowly and, because of differences in phasing, it easily results in differences between related dialects and languages. Especially related dialects tend to go through the same SC’s, but they don’t necessarily do so at the same rate and this is a cause of micro-variation. 3.2

Awareness

Both the productivity and the structural predictability of SCs decrease in accordance with the following cline:

(5) phonetic implementation > post-lexical process > lexical rule > lexicalized SC.

5. This is an inflected (2, 3 sg) verb form and /t/ is the inflectional ending. 6. In most varieties of spoken modern Dutch, fricatives tend to be realised voicelessly across the board, hence [s]out.

Chapter 2.  Of clocks, clouds and sound change 35



The speakers’ awareness (and consequently the manipulability) of a given phenomenon, however, increases along the same cline. Lexicalised SC is sometimes targeted by hypercorrection, the second type of opacity: forms which do not fit the structural description of a given rule do undergo / have undergone the rule. An example of correction in this sense is the split or unmerging of Seseo /s̪ / or Ceceo /θ/ in Andalusian Spanish in favour of the Castillian distinción /s̪ , θ/ (Villena-Ponsoda 2001; Regan 2017). Another example concerns lexicalised word-final [t] deletion (henceforth WFtD) in Afrikaans, a partly creolised former daughter language of Dutch. In nouns which have a final cluster of an obstruent followed by /t,d/,7 the /t,d/ only surfaces in the plural form, where it is followed by schwa, e.g. (6) a.

lig : ligte ‘light(s)’ hoof : hoofde ‘head(s)’

Some nouns have two plural forms, one with and one without the etymological plosive, e.g.  

b. kors : korste ~ korse ‘crust(s)’

Remarkably, in some nouns which do not have an etymological final plosive, the alternation occurs as well, e.g.  

c.

bos : bosse ~ boste ‘wood(s)’

while some nouns only have a plural form with a ‘hypercorrect’ /t/, such as8  

d. graf : grafte ‘grave(s)’ 8

Until 1925, Afrikaans was not recognized as a legitimate language in South Africa. Till then, standard Dutch (which does not allow WFtD) was the norm, but Afrikaans was the practice of language use. This fact probably underlies these hypercorrections. 4. Sociolinguistics Sociolinguistics has produced a number of highly relevant insights into the rise, embedding and evaluation of SC. One of the main contributions is Labov’s (1972) 13 stages model in which language and the speech community are intertwined in 7. Syllable-finally, the voice contrast neutralises into [t]. 8. The original Dutch plural form is gr/a/v-en. Gr/ɑ/f : gr/a/v-en is thus an instance of OSL, sketched in Section 3.1 above.

36

Frans Hinskens

the spread of SC.9 Three main steps in the 13 stage model have become known as indicators, markers and stereotypes; these notions refer to the specific status a variable phenomenon may have in (a sector of) the speech community. Unlike indicators, both markers and stereotypes show socio-stylistic variation and they are thus what Silverstein (2003) has labelled ‘enregistered’ and which are ‘available for social work’ (Johnstone 2006: 82). Unlike indicators and markers, stereotypes are commented on and often stigmatised. Cf. Table 2. Table 2.  Labov’s tripartition and two of the main distinguishing properties Property

Indicator = stage 2

Marker = stage 6

Stereotype = stage 12

socio-stylistic variation comment; stigmatization

no no

yes no

yes yes

Apart from a socially meaningful construct, style can also be a mechanical sideeffect of speech rate (e.g. Dressler’s 1975 allegro rules) in connected speech – and late phonological and phonetic processes such as reduction and assimilation tend to be style – specific. To return to Labov’s model, going from indicator via marker to stereotype, social meaning becomes more and more explicit. Social meaningfulness (or indexicality) presupposes awareness; the speakers tend not to be aware of indicators, but they are aware of markers and stereotypes. Socially stigmatised phenomena (changes which typically come up from below the level of conscious social awareness) can become prone to (hyper-) correction. An example of the latter mechanism is the replacement of the stigmatised palatal realisation of /n/, [ɲ], by the palatalised [nʲ] or Standard Modern Greek alveolar [n] variants in (C)ni(C) syllables, e.g. /paˈni/, ‘rag’, in the Patras variety of Greek (Pappas 2006). Speech communities can be cut up (both analytically and by the speakers themselves) in terms of macro-social dimensions such as social class, social mobility, age, gender, ethnicity etc. Also smaller scale, meso-social entities such as social networks and communities of practice have proven to be relevant for the spread of linguistic innovations. SC is usually transmitted in dense, multiplex social networks, where social cohesion is relatively big and where attitudes, norms and behavioural patterns tend to be relatively uniform. Two important mechanisms are incrementation and speaker agentivity. Incrementation results from the fact that, typically, every next generation slightly exaggerates the previous generation’s innovative pattern both quantitatively, i.e. in terms of the proportion of use of the new 9. In the variant presented in Labov 2001: 5: 517–518 more attention is paid to language acquisition / socialisation and perception.

Chapter 2.  Of clocks, clouds and sound change 37



variant, and qualitatively, which may result in rule generalisation (cf. Section 4.1 below). Agentivity stems from the fact that speakers do not always just react to e.g. linguistic innovations, but sometimes they initiate them. An example of the importance of the speakers’ attitudes and, at the same time, the indexicality of an innovation can be found in the work by Haddican et al. (2013) on the rate at which diphthongisation of the face and goat vowels spreads in York (UK). GOAT diphthongization

0.7

n=1

0.5

n=4

0.4

n=4 n=7 n=2

0.2 0.1 0.0

n=1

0.6 Euclidean distance

Euclidean distance

0.6

0.3

FACE diphthongization

0.7

n=4

0.5

n=4 n=7

0.4 0.3

n=2

0.2 0.1

0

1 2 3 4 Attitudinal index score

0.0

0

1 2 3 4 Attitudinal index score

Figure 2.  The relation between the proportion of use of new, diphthongised variants of two vowels and the speakers’ attitudes towards local community and dialect.10 From Haddican et al. (2013)

As the graphs in Figure 2 show, the stronger a speaker’s allegiance to the local community (measured on the horizontal axis), the more conservative, monophthongal variants s/he tends to use (vertical; Haddican 2013: 396). So the socio-emblematic meaning can slow down the tempo of the social embedding of a SC – or, in other cases, speed it up.

10. The Euclidian distance is the distance “between onset and offset using the first and ninth normalised values for F1 and F2” (p. 377); the higher the values the more diphthongal realisations. The bars headed by n = x indicate numbers of speakers.

38

Frans Hinskens

4.1

Exceptionlessness versus lexical diffuseness in the sociolinguistic study of sound change

Labov (2007) distinguished two different mechanisms for the spread of linguistic innovations, particularly SC. Transmission typically targets internally motivated, change; its vehicle is child L1 acquisition. The change is typically from below the level of conscious awareness, regular and it proceeds through multidimensional waves (cf. Schmidt’s 1872 Wellentheorie). Diffusion, on the other hand, typically targets change from above; here, linguistic change spreads through language contact and the vehicle is adult L2 acquisition – it proceeds in a lexically diffuse fashion, hence structurally unpredictable (“structural constraints are lost”, thus Labov 2007: 344). In the line of Labov’s proposal, there seems to be an older insight regarding the relationship between the internal and external, specifically geographical, extension of the transmission of SC, introduced by Robinson and van Coetsem (1973). According to this insight, in the area of origin a rule reaches eventually the most general application. An example concerns the palatalisation of /s/ in syllable onset across the Limburg dialects of Dutch studied by Goossens (1969); Goossens visualised his findings in the map in Figure 3 below. In the west-most area, numbered 1, palatal /ʃ/ only occurs in loans such as /ˈʃaʀəl/, ‘Charles’, and /mɑˈʃin/, ‘machine’. In area 2, it applies in loans and non-loans preceding vowels – and so forth. The rule has been generalised the most in area 6. In the classic generative phonological rule format, the original palatalisation rule A→B/C_D reached the largest geographical extension; during the process, the rule gradually was generalised internally into A→B/C_ (which underwent a more limited geographical spread) and subsequently maybe even into A→B, tout court, in the area of origin, here the area numbered 6. The simultaneous internal generalization and areal spread of a rule, as in the Robinson and Van Coetsem (1973) scenario, may result from incrementation (cf. above). Adding the stepwise internal growth as a dimension to the geographical coordinates, results in a three-dimensional picture which may bear resemblance to that of a cloud formation in Figure 4. Externally, every next wave of internal generalisation has been a bit less powerful and has expanded less. Applying the scenario to the /g/ → [j] change in non-standard varieties of German, sketched in Section 3.1 above, the first wave probably concerned the context preceding palatal vowels, the second wave concerned also non-palatal vowels and the third wave additionally concerned complex onsets, in which /g/ precedes a liquid /l, r/. Examples of diffusion are also available in the historical dialectology of Dutch. One of them concerns the two-step shift /uː/ > /yː/ > /œy/ discussed in Sections 2 and 3.1 above. The geographical and lexical distribution of these telescoped SCs of Middle and Early Modern Dutch have been studied in great detail by Kloeke

Chapter 2.  Of clocks, clouds and sound change 39



1 4

3 Hasselt

5 6 Maastricht

2

Aachen 2

4

Liège Distribution of phoneme /∫/ in Limburg

6

1 /∫/ in foreign words 2 /∫/ initially before vowels+1 3 /∫/ initially before /r/+1+2 4 /∫/ medially and finally+1+2+3 5 /∫/ initially before /l,m,n,p,t/+1+2+3+4 6 /∫/ initially before /w/ +1,2,3,4,5

Figure 3.  Distribution of /ʃ/ in a dialect cluster in Limburg (Goossens 1969)

Figure 4.  What a three-dimensional picture of the simultaneous internal and areal extension of a SC through transmission might look like

40 Frans Hinskens

(1927). Kloeke’s map (in Figure 5, in the version in Bloomfield 1933) clearly shows how, during its diffusion, a SC can affect different relevant items at different points in time, rather than simultaneously, thus resulting in lexical exceptions. The main exception here can be found in the group of dialects (in the outstretched northern and central eastern area) where the variants of “mouse” have retained the long high back vowel /u:/, while those for “house” have undergone palatalisation, yielding /y:/.

THE NETHERLANDS

Groningen

[mu:s, hu:s]

Leeuwarden

[mu:s, hy:s] [my:s, hy:s] [mф:s, hф:s] [mфys, hфys] The dotted area a Frisian Amsterdam Hague Rotterdam

Antwerp

Bruges Ghent

Brussels

R O M A N C E Figure 6. Distribution of syllabic sounds in the words mouse and house in the Netherlands.

— Afterapplies Kloeke. word-by-word; with respect to the number of affected items, Figure 5.  Diffusion the spread of a borrowed innovation often tapers off geographically and this is also visible in Kloeke’s map

The transmission-diffusion distinction occurred already in embryo in Schuchardt (1885: 76), where the author writes about “indigener versus verpflanzter Lautwandel”, endogenous vs. transplanted SC. There may be a third type. This option was discussed by the German dialectologist Penzl (1939) as Entlehnungslautwandel, ‘borrowed SC’, and Seidelmann (1972)

Chapter 2.  Of clocks, clouds and sound change 41



discussed it under the header of “rule borrowing”. The concept is comparable to Andersen’s (1988) “adaptive change”. Scheuringer (1992) claims that “real SC requires […] the absence of exemplary speech forms. […] Present-day SC is mostly motivated extra-linguistically through the orientation towards exemplary speech forms and it proceeds via sound substitution” (Lautersatz), i.e. lexically diffuse SC. But there is a “third possibility for SC, which has properties of both ‘real’ SC and sound substitution” (my translation, FH). Just like lexically diffuse SC, this third type of SC is prestige-driven and often boosted by its phonological “naturalness”; it has the latter property in common with regular Neogrammarian SC. The third mechanism of SC hence unites properties of the “classical” two. An example of a SC of this third type concerns the ongoing dismantling of the historical deletion of postvocalic /R/ preceding a coronal obstruent.11 It occurs in Ripuarian and East-Limburg dialects of Dutch, which are spoken in the southeastern corner of the language area. Examples are (7) a.

kort woord worst koorts eerst beurs baard

[kɔ̝t] [woːət] [wuːəʃ] [koːəts] [iˑəʃ] [byˑəʃ] [baːt]

‘short’ ‘word’ ‘sausage’ ‘fever’ ‘first’ ‘wallet’ ‘beard’

As far as the left-hand environment is concerned, the deletion has occurred after both tense and lax vowels, after back and front vowels, after rounded and unrounded vowels. In many Limburg dialects and in all Ripuarian dialects (including the subset of r-deleting ones), items such as eerst, beurs and worst have a high vowel; hence deletion occurred after high, mid and low vowels. As far as the righthand environment is concerned, the deletion took place preceding both voiced and voiceless stops, preceding fricatives and preceding both single and complex codas. Relatively recent loans such as  

b. mars [mɑʀʃ] sport [ʃpɔʀt]

‘march’ ‘sport’

never occur r-lessly, which shows that r-deletion has been lexicalised. To tell from apparent time data for the local Ripuarian dialect of Rimburg, R-lessness is undergoing loss, i.e. /r/ is being restored, resulting in convergence with prestigious standard Dutch and most other dialect varieties – but so far the 11. The remainder of this section is a paraphrase of parts of Hinskens 1992 Section 5.3.6.

42

Frans Hinskens

loss only occurs to a significant extent after short vowels, not after long vowels (F = 3.65 df = 2,24, p = .041). Thus R-lessness disappears first where it is needed least, namely in heavy (e.g. kort) rather than in superheavy syllables (e.g. woord) as, generally, super heavy syllables tend to be disprefererred. The dismantling of dialectal R-lessness is prestige-driven and to tell from the internal directionality, it is also boosted by phonological “naturalness”. 5. Cognitivist approaches Over the last decades a new paradigm has developed, which is sometimes referred to with the umbrella notion “cognitivist”. In connection with speech, sound structure, and SC, this includes Usage-based phonology (Bybee 2001; 2006; 2010) and Exemplar Theory (a.o. Johnson 1997; Pierrehumbert 2002). In this paradigm, grammatical knowledge is claimed to emerge bottom-up and structure is not “given a priori or by design” (Bybee 2010: 2). The frameworks do not assume faculties such as the language acquisition device, nor a mental blue-print for grammar, nor a language bioprogram. “The cognitive and psychological processes and principles that govern language are not specific to language, but are in general the same as those that govern other aspects of human cognitive and social behavior” (Bybee 2001: 17). Language and linguistic behaviour embedded in “domain general” cognitive skills and processes. In this approach language use requires little, if any computation. In principle everything is stored redundantly: every realisation of every item is stored in anecdotal memory in bundles of maximally concrete articulatory, acoustic, grammatical, semantic and pragmatic information concerning the single occurrences (“tokens” or “exemplars”). All occurrences are organised in exemplar clouds, around a prototype; the lexicon is a cosmos of multi-dimensional networks. Token frequency (i.e. frequency of usage) and type frequency (distributional frequency) are the stem cells of grammar. Closely related to type frequency is the concept of “neighbours” (Luce and Pisoni 1998), i.e. words which have a similar phonological shape, “words that are one sound away from a given word”; similarly, transitional probabilities, e.g. “the probability of a C given a final V” (Coetzee 2008: 250). A related concept is the informativity of a sound, i.e. its average predictability across all segmental contexts (Cohen Priva 2017). In this view, the productivity of a phenomenon is influenced by its type frequency. In connection with token frequency mention should be made of local frequency manifestations such as “recency” / “audience design” effects (e.g. Schwarz 2012), an instantiation of priming effects. The Information-theoretic tools of probability, entropy and surprisal value (Zampaolo 2016) are also based on token frequency.



Chapter 2.  Of clocks, clouds and sound change 43

It has been claimed that the more frequently a word is used, i.e. the higher its token frequency, the more predictable it is. And the more predictable, the more easily it can be phonetically reduced.12 On the other hand, with respect to morphophonology, “analogical levelling affects low frequency words before high frequency words […] sometimes does not affect high frequency forms at all” (Bybee 2012: 216, 225). In this paradigm SC is always both gradient and lexically diffuse, since there is no such concept as the segment. A rule or process can be partly lexicalised or morphologised, hence lexically stored, while it is still productive; cf. Bermúdez-Otero’s Stratal Optimality Theoretical concept of “rule scattering”, through which “a process in one component of the grammar gives rise to a new rule at a higher level […] without ceasing to apply at a lower level” (Bermúdez-Otero 2015: 2). With respect to phonetic change, an important question is whether token frequency is merely relevant to reductive change (lenition, deletion, assimilation and the like) or to any type of SC, including fortition and the like (e.g. chain shifting); the latter position is held by Pierrehumbert (2002). The changes West Germanic13 (henceforth Wgm) /u:/ underwent in the history of Dutch are instances of fortition. Van Reenen and Elias (1998: 108) present relevant data for 18 relevant items in the 353 Dutch dialects. They do not, however, take the logically following step of relating stability versus change of Wgm /u:/ with the items’ token frequencies. The Pearson r for the number of dialects in which /u:/ has been changed and the frequency of usage appears to be .294 (two-tailed p = .236) which suggests that token frequency may not have played a role in the fortition type of SCs that the vowel underwent. However, when the token frequencies are log-transformed14 a significant yet moderate correlation is found (r =.503, two-tailed p >.05); the frequency of usage explains some 25% of the variance in the number of dialects in which /u:/ has been changed. The outcome does corroborate Kiparsky’s (1988: 371) and Kerswill’s (2011: 229) claim that token frequency does not usually play a role in connection with geographical spread. However, as the data concern historical change, these findings might not be conclusive. Finding indications for lexical effects in historical sound change on the basis of modern frequency data may be even more difficult than reconstructing astronomical phenomena that took place light-years away using data from cosmic background radiation. 12. Almeida (2013) discusses the relationship between token frequency and phonetic alternation in the realisation of /t͜ ʃ/and intervocalic /d/ in the Santa Cruz de Tenerife variety of Canarian Spanish, concluding that it is conceivable that the frequency effects can be modified by other variables. 13. The ancestor of present-day English, Scots, Frisian, German and Dutch. 14. Logarithmic token frequencies were calculated as log(token frequency + 1). Using this formula, a token frequency of 0 remains 0.

44 Frans Hinskens

Table 3.  The retention of Wgm /u:/ in 18 items and the items’ token frequencies (after data from Van Reenen and Elias 1998: 108) Modern st.Dutch

English gloss

N of dialects in which /u:/ has been changed

duif vuist muis duim kruipen ruit buik uil bruin zuigen duizend buiten huis uit gebruiken luisteren luiden kuip

‘dove’ ‘fist’ ‘mouse’ ‘thumb’ ‘to creep’ ‘window-pane’ ‘belly’ ‘owl’ ‘brown’ ‘to suck’ ‘thousand’ ‘outside’ ‘house’ ‘out’ ‘use’ ‘to listen’ ‘sound’ ‘tub’

201 211 212 216 221 228 237 241 242 246 275 276 282 290 300 325 328 328

Frequency of oral usage (in two corpora of modern spoken Dutch)   1   1   0   2   8   4   4   0   6   5  17 146 372 658  55  38   2   3

6. Towards an integrated theory SC is obviously too complex a phenomenon to be considered as the domain of one single branch of linguistics. According to Kiparsky (1988: 399), “the same feature, indeed the same rule, [can] be subject to lexical diffusion in one language or dialect and not in another depending on whether the feature is lexically distinctive or not”. An example of a SC which has a different rule typological status in related language systems is WFtD; whereas in the Nijmegen urban dialect of Dutch it is a post-lexical process, in many Limburg dialects of Dutch it applies post-lexically and lexically15 and in Afrikaans it has been lexicalized. A theory is needed which can (1) account for lexical frequency effects with regard to productive processes of phonetic reduction (lenition and the like) and in the course of a SC’s lexicalization; (2) accommodate the insights into the life cycle 15. At least in the dialects of Ubach over Worms and Rimburg, studied in Hinskens (1992), the rule is also lexical since it is statistically sensitive to morphological structure. Cf. Guy’s (1991) ‘exponential effects’.

Chapter 2.  Of clocks, clouds and sound change 45



of SC. The way the theory has been embedded in formal phonological theory (very transparently so by Ramsammy 2015), however, seems to apply only to internally generated SC; (3) account for the diffusion of exogenous SC. The prestige-driven borrowing of SC (Labov’s 2007 diffusion) typically enters the grammar in the lexicalisation phase. In Table 4 the three perspectives and the properties they ascribe to the various types of SC are aligned. Each status of a SC in the rule typology corresponds to a specific set of internal, cognitive and social properties. The combination of properties in three different domains adds dynamics to the constellation.

lexicalization

stem

lexical

word

post-lexical

phrase

phonetic

minimal ↑

↓ maximal

Social style effect

Cognitive

status in Labov’s tripartition

conditioning

productivity

grammatical level

rule typology

Internal

lexical frequency effects

Table 4.  The three perspectives and the properties they ascribe to the various types of SC

lexical

token and type freq. effects on lexicaliz. of all types of processes

stereotype or marker

social

grammatical

none

not applicable

none

marker or indicator

social and/or mechanical

indicator

mechanical

prosodic or phonological unconditioned or phonetically conditioned

token freq. effects only on reduction and similar processes

From this integrated theory, a number of empirically testable hypotheses can be derived, including: 1. Post-lexical processes can be rudimentarily lexicalised; 2. Usage-related factors only play a role where grammar underdetermines usage (cf. Anttila 2006); 3. Frequency of usage only affects lenition, reduction, assimilation and the like and it does not affect fortition-type phenomena; 4. “High frequency words and phrases undergo phonetic reduction at a faster rate than low- and mid-frequency sequences” (Bybee 2006: 714); 5. There is an inverse relation between the productivity of a phenomenon and the degree to which the speakers are aware of it;

46 Frans Hinskens

6. Indicators are typically phonetic rules and postlexical processes, while stereotyping and hypercorrection will typically involve lexically diffuse SC; 7. The layered organization of SC adds coherence: the closer different SCs are rule-typologically, the more similar they will be in terms of internal, social and cognitive organisation. For most of these hypotheses, some evidence and/or some methodological considerations will be briefly sketched below; needless to add that more systematic research is called for. Re hypothesis 1: post-lexical processes can be rudimentarily lexicalised. This can be exemplified with word-final [t] deletion in the East-Limburg dialect of Ubach over Worms; in this local dialect (and many other Limburg dialects), WFtD is fully productive. Nevertheless there are words such as the verbs (8) kɔstə ~ kɔsə mɛldə ~ mɛlə

‘to cost’ ‘to announce’; ‘to register’

where the deletion has been lexicalised. Re hypothesis 2: usage-related factors only play a role where grammar underdetermines usage (cf. Anttila 2006). In modern spoken standard Dutch, unstressed full vowels can be reduced to schwa or – depending on phonotaxis – even to deletion. Interestingly, there are structurally similar items which show dramatical differences in this respect, such as e.g. (9) m/i/nuut ~ m[ə]nuut ‘minute’ but   p/i/loot ~ *p[ə]loot ‘pilot’

Segmentally, prosodically and stress-wise these words are very similar; yet the vowel in the unstressed first syllable only undergoes reduction in m/i/nuut. This is clearly a case where grammar underdetermines usage. In connection with this reduction process, there are obviously more determinants than formal structure alone. Re hypothesis 3: frequency of usage only affects lenition (reduction, assimilation and the like) and it does not affect fortition-type phenomena. On the basis of F2 measurements, Dinkin (2008) studied the centralisation of short vowels (an instance of lenition) in dialects of English spoken in the northeast of the USA; it appeared that high-frequency words are more subject to centralisation. The outcomes of the analyses of data regarding Wgm /uː/, presented at the end of Section 5 above, seem to suggest that token frequency does not necessarily affect fortition. Re hypothesis 4: “high frequency words and phrases undergo phonetic reduction at a faster rate than low- and mid-frequency sequences” (Bybee 2006: 714). This implies that high and low frequency items have different slopes. In the data on



Chapter 2.  Of clocks, clouds and sound change 47

co-articulatory vowel nasalisation in Philadelphia studied by Zellou and Tamminga (2014), however, the effects of the speakers’ birth-year and token frequency are independent, showing that high frequency words on the one hand and low- and mid-frequency sequences, on the other, do not have different slopes. This finding is reminiscent of Kroch’s (1989) “constant rate effect”. Re hypothesis 5: there is an inverse relation between the productivity of a phenomenon and the degree to which speakers are aware of it. Phonetic phenomena tend to be automatised, hence fully productive and below the level of social awareness. On the other hand, phenomena which have risen so high above the level of social awareness that they have become stigmatised are at best prone to lexically diffuse spread (cf. Labov 2006: 509). Re hypothesis 6: indicators are typically phonetic rules and postlexical processes, while stereotyping and hypercorrection will typically involve lexically diffuse SC. Operationally, this hypothesis can be translated into two related claims: (1) for a process to be subject to style shifting it needs to be at least post-lexical; (2) (hyper-) correction usually targets lexicalising or lexicalised processes. Evidence for the second part of the claim includes etymologically hypercorrect plurals in Afrikaans, discussed in Section 3.2 above. An interesting source of evidence is the type of hyperdialectisms introduced by non-native (L2) speakers or semi-speakers, who sometimes over-apply a dialect feature in contexts where it does not “belong” historically. These speakers “do not know any better: their analysis of the target variety is faulty” and for that reason they extend a given phenomenon “into words where it is not historically justified” (Trudgill 1988: 551,553). Realizing the East-Limburg or Ripuarian dialect variants of items such as mars, ‘march’, and sport, ‘sport’, R-lessly (Section 4.1 above) would be an instance of hyperdialectism. Hyperdialectisms of this type are common in make-shift dialect varieties. An example (from Swanenberg 2009) is diminutive clubske for traditionally well-formed clubke, standard Dutch clubje, ‘little club’. This variant has a Brabantish16 ring to it and it may well have been used to flag a (claim to) Brabant identity of the speaker. However, the authentic dialect variant of the diminutive would not have the -skə allomorph, which merely occurs following stems ending in a velar.17 Re hypothesis 7: the closer different SCs are rule-typologically, the more coherent they are, i.e. the more similar they are in terms of internal, social and cognitive organisation. And, logically, the bigger the rule typological distance between different SCs, the less coherent they are. This claim can be tested for different SCs within

16. Or Limburg, for that matter. 17. See Hinskens (2004: 17–19 and 2014: 114–116) for more considerations and examples.

48 Frans Hinskens

the same language variety and for a SC which has a different rule typological status in different related language varieties. Future work can consist of interpreting available studies against the background of the present proposal and designing new empirical studies to systematically test all hypotheses.

References Almeida, Manuel. 2013. La frecuencia de las palabras en los proceses de variación y cambio. Revista de la Sociedad Española de Lingüística (RSEL) 43 (2). 37–62. Andersen, Henning. 1988. Center and periphery: Adoption, diffusion and spread. In J. Fysiak (ed.), Historical dialectology: Regional and social, 39–83. Berlin: de Gruyter. https://doi.org/10.1515/9783110848137.39 Anttila, Arto. 2006. Variation and opacity. Natural Language and Linguistic Theory 24. 893–944. https://doi.org/10.1007/s11049-006-0002-6 Bermúdez-Otero, Ricardo. 2015. Amphichronic explanation and the life cycle of phonological processes. In P. Honeybone and J. Salmons (eds.), The Oxford handbook of historical phonology, 374–399. Oxford: Oxford University Press. Bloomfield, Leonard. 1933. Language. London: Unwin. Browman, Catherine and Louis Goldstein. 1992. Articulatory phonology: An overview. Phonetica 49. 155–180.  https://doi.org/10.1159/000261913 Bybee, Joan. 2001. Phonology and language use. Cambridge: Cambridge University Press. https://doi.org/10.1017/CBO9780511612886 Bybee, Joan. 2006. From usage to grammar: The mind’s response to repetition. Language 82 (4). 711–733.  https://doi.org/10.1353/lan.2006.0186 Bybee, Joan. 2010. Language, use and cognition. Cambridge: Cambridge University Press. https://doi.org/10.1017/CBO9780511750526 Bybee, Joan. 2012. Patterns of lexical diffusion and articulatory motivation for sound change. In M. J. Solé and D. Recasens (eds.), The initiation of sound change: Perception, production, and social factors, 211–234. Amsterdam-Philadelphia: John Benjamins Publishing. https://doi.org/10.1075/cilt.323.16byb Chomsky, Noam and Morris Halle. 1968. The sound pattern of English. New York: Harper and Row. Coetzee, Andries. 2008. Grammaticality and ungrammaticality in phonology. Language 84 (2). 218–257.  https://doi.org/10.1353/lan.0.0000 Cohen Priva, Uriel. 2017. Informativity and the actuation of lenition. Language 93 (3). 569–597. https://doi.org/10.1353/lan.2017.0037 Dinkin, Aron. 2008. The real effect of word frequency on phonetic variation. Penn Working Papers in Linguistics 14 (1). 97–106. Dressler, Wolfgang. 1975. Methodisches zu Allegro-Regeln. In W. Dressler et  al. (eds.), Phonologica 1972, 219–234. München: Fink. Garrett, Andrew. 2009. Patterns of sound change. Slides for a class in a course on Historical Linguistics in the 2009 Linguistic Institute of the Linguistic Society of America, held at the University of California, Berkeley.



Chapter 2.  Of clocks, clouds and sound change 49

Goossens, Jan. 1969. Strukturelle Sprachgeographie. Eine Einführung in Methodik und Ergebnisse. Heildelberg: Winter. Guy, Gregory. 1991. Explanation in variable phonology: An exponential model of morphological constraints. Language Variation and Change 3. 1–22. https://doi.org/10.1017/S0954394500000429 Haddican, Bill, Foulkes, Paul, Vincent Hughes and Hazel Richard. 2013. Interaction of social and linguistic constraints on two vowel changes in northern England. Language Variation and Change 25 (3). 371–401.  https://doi.org/10.1017/S0954394513000197 Hinskens, Frans. 1992. Dialect levelling in Limburg. Structural and sociolinguistic aspects. PhD-thesis University of Nijmegen (revised and abridged version 1996, Niemeyer, Tübingen). Hinskens, Frans. 2004. Nieuwe regenboogkleuren. Jonge typen niet-standaardtaal en hun taalkundig belang. Inaugural address, Vrije Universiteit Amsterdam. Retrieved from http://www. meertens.knaw.nl/medewerkers/frans.hinskens/rede.pdf. Hinskens, Frans. 2014. Despite or because of intensive contact? Internal, external and extralinguistic aspects of divergence in modern dialects and ethnolects of Dutch. In K. Braunmüller, S. Höder, and K. Kühl (eds.), Stability and divergence in language contact: Factors and mechanisms, 109–140. Amsterdam-Philadelphia: John Benjamins Publishing. Hinskens, Frans. 2020. The expanding universe of the study of sound change. To appear in R. Janda, B. Joseph and B. Vance (eds.), The handbook of historical linguistics, 2nd ed. Hoboken-Oxford: Wiley Blackwell. Janda, Richard. 2003. ‘Phonologization’ as the start of dephonetization – or, on sound change and its aftermath: Of extension, generalization, lexicalization, and morphologization. In B. Joseph and R. Jansa (eds.), The handbook of historical linguistics, 410–422. Malden MA-Oxford: Blackwell.  https://doi.org/10.1002/9780470756393.ch9 Janda, Richard and Brian Joseph. 2003. On language, change and language change – or, of history, linguistics, and historical linguistics. In B. Joseph and R. Janda (eds.), The handbook of historical linguistics, 3–180. Malden MA-Oxford: Blackwell. https://doi.org/10.1002/9780470756393.ch Johnson, Keith. 1997. Speech perception without speaker normalization: An exemplar model. In K. Johnson, J. Mullennix (eds.), Talker variability in speech processing, 145–165. San Diego: Academic Press. Johnstone, Barbara, Jennifer Andrus and Andrew Danielson. 2006. Mobility, indexicality, and the enregisterment of “Pittsburghese”. Journal of English Linguistics 34 (2). 77–104. https://doi.org/10.1177/0075424206290692 Joseph, Brian. 2012. Lexical diffusion and the regular transmission of language change in its sociohistorical context. In J. Hernández-Campoy and J. Conde-Silvestre (eds.), The handbook of historical sociolinguistics, 408–426. Malden USA: Blackwell. https://doi.org/10.1002/9781118257227.ch22 Kerswill, Paul. 2011. Sociolinguistic approaches to language change: Phonology. In R. Wodak, B. Johnstone and P. Kerswill (eds.), The Sage handbook of sociolinguistics, 219–235. London: Sage.  https://doi.org/10.4135/9781446200957.n17 Kiparsky, Paul. 1968. Linguistic universals and linguistic change. In E. Bach and R. T. Harms (eds.), Universals in linguistic theory, 171–210. New York: Holt, Rinehart and Winston. Kiparsky, Paul. 1988. Phonological change. In F. Newmeyer (ed.), Linguistics: The Cambridge survey, vol. 1, Linguistic theory: Foundations, 363–413. Cambridge: Cambridge University Press.

50

Frans Hinskens

Kiparsky, Paul. 1995. The phonological basis of sound change. In J. Goldsmith (ed.), The handbook of phonological theory, 640–670. Cambridge, Mass: Blackwell. Kloeke, Gesinus. 1927. De Hollandsche expansie in de zestiende en zeventiende eeuw en haar weerspiegeling in de hedendaagsche Nederlandsche dialecten. Den Haag: Nijhoff. Kroch, Anthony. 1989. Reflexes of grammar in patterns of language change. Language Variation and Change (1/3). 199–244.  https://doi.org/10.1017/S0954394500000168 Labov, William. 1972. Sociolinguistic patterns. Philadelphia: University of Pennsylvania Press. Labov, William. 1981. Resolving the neogrammarian controversy. Language 57 (2). 267–308. https://doi.org/10.2307/413692 Labov, William. 2001. Principles of linguistic change. Volume 2: Social factors. Oxford: Blackwell. Labov, William. 2006. A sociolinguistic perspective on sociophonetic research. Journal of Phonetics (34). 500–515.  https://doi.org/10.1016/j.wocn.2006.05.002 Labov, William. 2007. Transmission and diffusion. Language 83 (2). 344–387. https://doi.org/10.1353/lan.2007.0082 Luce, Paul and David Pisoni. 1998. Recognizing spoken words. The neighborhood activation model. Ear and hearing 19. 1–36.  https://doi.org/10.1097/00003446-199802000-00001 Moulton, William. 1961. Lautwandel durch innere Kausalität: Die Ostschweizerische Vokal­ spaltung. Zeitschrift für Mundartforschung, 28. 227–251. Ohala, John. 2003. Phonetics and historical phonology. In B. Joseph and R. Janda (eds.), The handbook of historical linguistics, 669–686. Malden MA-Oxford: Blackwell. https://doi.org/10.1002/9780470756393.ch22 Osthoff, Hermann and Karl Brugmann. 1878. Morphologische Untersuchungen auf dem Gebiete der indogermanischen Sprachen. Leipzig: Hirzel Pappas, Panayiotis. 2006. Stereotypes and /n/ variation in Patra, Greece. Results from a pilot study. In F. Hinskens (ed.), Language variation – European perspective, 153–167. AmsterdamPhiladelphia: John Benjamins Publishing.  https://doi.org/10.1075/silv.1.11pap Penzl, Herbert. 1939. Kompromissvokal‘ und Lautwandel. Anglia 63. 88–99. Pierrehumbert, Janet. 2002. Word-specific phonetics. In C. Gussenhoven and N. Warner (eds.), Laboratory phonology VII, 101–140. Berlin: de Gruyter. https://doi.org/10.1515/9783110197105.101 Popper, Karl. 1966. Of clocks and clouds. An approach to the problem of rationality and the freedom of man. St. Louis: Washington University Ramsammy, Michael. 2015. The life cycle of phonological processes: Accounting for dialectal microtypologies. Language and Linguistics Compass (9/1). 33–54. https://doi.org/10.1111/lnc3.12102 Regan, Brendan. 2017. The effect of dialect contact and social identity on fricative demerger. PhDthesis, University of Texas at Austin. Robinson, Orrin and Frans van Coetsem. 1973. Review article of R. King, 1969, Historical linguistics and generative grammar. Lingua 31. 331–369. https://doi.org/10.1016/0024-3841(73)90031-4 Scheuringer, Hermann. 1992. Die grammatikalische Erklärung von Sprachvariation. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikations-forschung. (ZPSK), 45 (5). 481–494. Scheutz, Hannes. 2005. Aktuell stattfindender Lautwandel/Sound change in progress. In U. Ammon, N. Dittmar and K. J. Mattheier (eds.), An international handbook of the science of language and society. Sociolinguistics, 2nd ed., 1704–1717. Vol. 2. Berlin-NewYork: de Gruyter.



Chapter 2.  Of clocks, clouds and sound change 51

Schleicher, August. 1861. Compendium der vergleichenden grammatik der indogermanischen spra­chen. Weimar: Böhlau. Schmidt, Johannes. 1872. Die Verwandtschaftsverhältnisse der indogermanischen Sprachen. Weimar: Böhlau. Schuchardt, H. 1885. Über die Lautgesetze: Gegen die Junggrammatiker. In Leo Spitzer (ed.), Hugo Schuchardt-Brevier. Ein Vademekum der allgemeinen Sprachwissenschaft, 51–87. Niemeyer, Halle. Schwarz, Christoph. 2012. Rezenz als Faktor phonologischer Variation. Zeitschrift für Literatur­ wissenschaft und Linguistik (LiLi) 166. 29–44.  https://doi.org/10.1007/BF03379862 Seidelmann, Erich. 1972. Lautwandel, Lautersatz und die Bedingungen des Sprachwandels. In A. Weiss (ed.), Dialekte im Wandel, 111–127. Göppingen: Göppinger Arbeiten zur Linguistik, 538. Silverstein, Michael. 2003. Indexical order and the dialectics of sociolinguistic life. Language and Com­munication 23. 193–229.  https://doi.org/10.1016/S0271-5309(03)00013-2 Swanenberg, Jos. 2009. Van alterande sorte. Brabants tussen dialect en standaardtaal. Tilburg: Universiteit van Tilburg. Trudgill, Peter. 1988. On the role of dialect contact and interdialect in linguistic change. In J. Fisiak (ed.), Historical dialectology. Regional and social, 547–563. Berlin: de Gruyter. https://doi.org/10.1515/9783110848137.547 Van Reenen, Pieter and Michael Elias. 1998. Taalverschillen. Een werkboek over variatie en verandering in taal. Bussum: Coutinho. Villena-Ponsoda, Juan-Andrés. 2001. La continuidad del cambio lingüístico. Granada: Univer­si­dad. Wang, William. 1969. Competing changes as a cause of residue. Language 45. 9–25. https://doi.org/10.2307/411748 Zampaolo, André. 2016. Sibilant sound change in the history of Portuguese. An Informationtheoretic approach. Diachronica 33 (4). 503–529. https://doi.org/10.1075/dia.33.4.03zam Zellou, Georgia and Meredith Tamminga. 2014. Nasal coarticulation changes over time in Philadelphia English. Journal of Phonetics, 47. 18–35.  https://doi.org/10.1016/j.wocn.2014.09.002

Chapter 3

Evaluations of foreign accent in a purist speech community The case of Iceland Stefanie Bade

University of Iceland

This study was directed both at evaluations of foreign-accented Icelandic according to listeners’ background and the possible consequences of evaluations for the linguistic climate in Iceland. Seven female non-native speakers, representing the largest immigrant groups, and one native speaker of Icelandic were recorded and subsequently evaluated according to eight personality traits. Statistical analysis of answers of 538 Icelanders revealed that women and those over 60 are generally more positive in their evaluation as compared to men or those under 60 years. Overall, results imply, firstly, that those accents which can be categorised as Western are preferred over those that belong to Eastern Europeans or Asians and, secondly, that the native Icelandic accent is not always rated highest. Keywords: Icelandic, foreign accent, language attitudes, stereotypes, language variation, verbal guise, listener background, linguistic climate

1. Introduction Iceland has been a culturally and linguistically homogeneous country since its settlement over 1000 years ago. The country’s unique linguistic climate has, however, lately been contested by an ever-growing number of non-native speakers that speak Icelandic with a foreign accent, now amounting to 12% of the population. It is, thus, conceivable that another evaluation system along native/non-native dimensions is emerging. In order to gain an insight into attitudes towards foreign-accented Icelandic, the following study1 has three aims. The first is to establish whether evaluations of 1. This study is funded by the Icelandic Research Fund (Rannís) and the Research Fund of the University of Iceland (Rannsóknarsjóður Háskóla Íslands). https://doi.org/10.1075/silv.22.03bad © 2019 John Benjamins Publishing Company

54

Stefanie Bade

non-native speech can cause a possible shift in the evaluation system that is traditionally built on ideas of standardness/non-standardness, thereby making way for a different or additional evaluation system based on differentiation of nativeness/ non-nativeness. The second is to examine whether there are differences in evaluating foreign-accented Icelandic that can be attributed to certain nationalities and/ or ethnic backgrounds. The third is to ascertain whether the background variables gender and age exercise an influence on the evaluation of foreign-accented speech. A verbal guise experiment (Garrett 2010: 41–42) using eight female speakers, seven non-native and one native Icelandic speaker were presented to native speakers of Icelandic for evaluation on differential semantic scales. In order to obscure the intent of the study, the experiment was introduced by a cover story that made the speakers participants in employment interviews, as applicants for a job in a multicultural playschool. 2. Evaluating an accent Of all the components that make up speech, phonological features are said to be especially salient (Moyer 2013: 85). Clues of only a few seconds (Kang and Rubin 2009: 2; Moyer 2013: 85) give away a person’s language background. Therefore, perceiving a non-native accent is inevitably intertwined with certain ideas about the speaker and his supposed character traits, whether they have a positive or negative nature (Giles and Niedzielski 1998). The idea of a distinction between native and non-native speech is thus not solely built on different phonological features, but on a sociopsychological distinction between adhering to and deviating from an established standard. Likewise, the examination of foreign-accented speech is closely connected with social stereotypes that are subconsciously triggered by the perception of linguistic anomalies. Extensive research on reactions to non-native speech confirms that accents are associated with social/economic values, such as correctness, desirability, prestige, and power. These values are accessible by employing adjectives presented in semantic differential scales (see e.g. Brennan and Brennan 1981; Kristiansen 2006; Ladegaard 1998; Ryan and Carranza 1975; Ryan and Sebastian 1980). Traditionally, studies of attitudes towards varieties of spoken language have used the matched-guise technique (see Lambert et al. 1960), which is used in research aimed at investigating isolated accents as triggers for perceptions of personal traits. In order to fit the purpose of this study verbal guises were used, in which audio stimuli were recorded by different speakers. When assessing different pronunciations, double standards seem to apply depending on the nature of the utterance. In contrast to a most salient standard/ non-standard dichotomy when assessing native accents (cf. Cargile and Giles 1998),



Chapter 3.  Evaluations of foreign-accented Icelandic 55

“the distinction native-non-native is front and center in our minds […] [w]hen faced with L2 speakers” (Moyer 2013: 102, original italics). In terms of a standard/ non-standard dichotomy Cargile and Giles point out that “most accents can be classified by the degree to which they are considered ‘standard’ or ‘non-standard’ within a particular community” (1998: 340). Classifications of this kind are built on speakers’ ideas of what is considered good (standard) and bad (non-standard) language within the spectrum of a language native accents. Those ideas are commonly accepted and reproduced by the speech community, thus constituting a strong element in a speech community’s language ideology. In this regard, speakers of non-standard accents are evaluated differently (and usually less favourably) than those that speak with a standard accent, especially in terms of status-related issues. In contrast, a differerent kind of evaluation system seems to apply when faced with non-native accents, given that foreign accents deviate from a native (standard) accent, thus embodying a non-standardness that operates outside the system in which native accents are evaluated. In the case of Iceland, where language purism rooted in beliefs about correctness and pure language constitutes an undisputed element in public discourse, it can be hypothesised that this former type of dichotomy has been valid for centuries in terms of native Icelandic. Recent immigration, growing global interconnectedness as well as a resulting change in social and linguistic circumstances might, however, bring about an (additional) evaluation system, built on a native/non-native distinction. Since it is expected that such a system would operate outside the spectrum of native Icelandic with native Icelandic at the same time representing the favoured variety, deep-rooted language ideological concepts should remain untouched. 3. The language situation in Iceland Geographic isolation, a strong idiosyncratic literary tradition as part of cultural history, the struggle for independence from the Danish crown in the 19th century as well as a conservative and protectionist language policy have substantially contributed to Iceland’s unique linguistic climate. This has given rise to a standard variety that the general public holds in high regard (Leonard and Árnason 2012: 92; Hilmarsson-Dunn and Kristinsson 2010). This standard variety is supposed to be pure and free of foreign influences (cf. Leonard and Árnason 2012: 94). At the same time, it automatically constitutes an ideal which results in beliefs about good and bad language and language use, echoing the standard/non-standard dichotomy outlined above. A pilot study with two focus groups indicates that the younger generation in Iceland tends to be more tolerant towards foreign influences (Bade and Isenmann

56

Stefanie Bade

unpublished; see also Árnason 2003). The matched-guise part of the MIN (Moderne Importord i Språka i Norden) project’s investigation of attitudes towards English in the North, has revealed a high level of “linguistic consciousness” (Ewen and Kristiansen 2006; Leonard and Árnason 2012: 9; Óladóttir 2009) among the Icelandic public, confirming a general lay interest in as well as a firm purist attitude towards the Icelandic mother tongue. Thus, the importance of the Icelandic standard variety, especially as primary marker for Icelandic national identity, does not seem to decline (cf. Kristiansen and Vikør 2006). In the case of Icelandic pronunciation, there has been little variation and it has been largely agreed that Icelandic as a whole is a rather homogenous language in comparison to others (Árnason 2005: 366). The phonological variation that exists in Iceland today is not largely subject to social evaluation, although research indicates that some varieties are considered more attractive than others (Hlynsdóttir 2016: 29). While those phonological differences are naturally regarded as home-grown Icelandic, the role of foreign accents and their evaluation has not yet been investigated. The analysis of the gender variable within the matched-guise part of the Icelandic MIN project (Árnason 2006: 23; Ewen and Kristiansen 2006: 43) confirmed assumptions that women tend to prefer standard forms over men (cf. Adamson and Regan 1991). With regard to the variable age, results of the project showed that the two middle age groups of those between 30–59 evaluate the English guise more negatively than the youngest (18–29) and oldest age groups (60+). 4. Method 4.1

Speakers

Speakers in this study represent the seven largest immigrant groups into Iceland, with an omission of speakers of Swedish and British English2 (see Figure 1), and come from the following countries: Denmark, Germany, Lithuania, the Philippines, Poland, Thailand, and the USA. Additionally, one native speaker of Icelandic was chosen to produce a verbal guise.

2. Speakers of Swedish and British English were excluded for several reasons. Firstly, it was attempted to achieve a relatively equal representation of geographical areas. This is why Danish was chosen over Swedish with the common Icelandic-Danish history as contributing factor. US-American English was selected over British English since no suitable speaker was found that complied with selection criteria (see this chapter). Secondly, it seemed reasonable to limit the number of recordings used in this study in order to prevent fatigue of participants.



Chapter 3.  Evaluations of foreign-accented Icelandic 57

In order to recruit non-native speakers from various backgrounds all with a similar level of proficiency in Icelandic, students of Icelandic as a Second Language at the University of Iceland were contacted. Students willing to participate were pre-categorised according to age, gender, and country of origin/mother tongue. Subsequently, a questionnaire was sent to them, exploring criteria potentially influencing degree of accent: length of residence, length of L2 learning, use of L1 and L2, and motivation (see Flege, Munro and MacKay 1995; Purcell and Suter 1980). Finally, only female students in their second year within the programme Icelandic as a Second Language that were between 25–35, had the requested country of origin/mother tongue and length of L2 learning were eventually chosen for the recordings.3 Age and gender of speakers were specified in order to reflect the main background criteria of playschool staff in Iceland, i.e. the group that formed the basis of the study as it was presented to the participants in order to obscure the actual intent (see 4.2). The native Icelandic speaker was recruited according to the same criteria for gender and age. The text of choice is an oral exam test used within the programme Icelandic as a Second Language at the University of Iceland and comprises most of the sounds occurring in Icelandic. It is a short story about a cat and her adventures, suitable for small children at pre-school age. It was sent to the speakers prior to the recording sessions in order to give them ample time to familiarise themselves with the text and potentially prevent disruptions in reading fluency.4 Audio stimuli were produced by each speaker with each recording containing between 58 and 88 seconds of read text. As one perhaps might expect, the recording of the native speaker of Icelandic is the shortest with 58 seconds. Considering that studies have found speaker disfluency to be a factor in evaluations of non-native accents (see e.g. Ryan 1983; White and Lee 1991), a possible influence of individual speech rate and/or fluency on evaluations of the voices in this study cannot be excluded.5 3. Although speakers were carefully chosen with regard to the criteria mentioned, bias of results deriving from differences in relative competences cannot be excluded. 4. Employing the verbal-guise technique can have certain methodological implications. Firstly, reading a text aloud gives rise to a somewhat artificial situation possibly influencing especially suprasegmental features such as speech rate, intonation and stress. Secondly, individual accent features of the seven non-native speakers, including L1 to L2 transfer of suprasegmental features and individual ability to mimic an Icelandic accent, can play a distinctive role in speaker performance and listener perception. Thirdly, apart from individual voice quality, some voices may simply sound more attractive, relaxed, etc. than others, whatever the language may be. 5. An example of a possible influence of speech fluency on evaluations is the native speaker of Icelandic considerably being downgraded on the trait efficient, while receiving the highest score for the trait relaxed (see Figure 3).

58

Stefanie Bade

16000 13811

14000 12000 10000 8000 6000 4000 2000

3412 1132

2001

1757

0 Denmark Lithuania Sweden Germany

2183

1489 Britain

Poland

USA

1727

1279

Philippines Thailand

Figure 1.  The nine largest immigrant groups in Iceland acc. to Statistics Iceland, 1.1.2017a

4.2

Procedure

Data collection was conducted from the 17th of May until the 20th of August, 2016. A stratified random sample according to gender, age and residency, was taken among 1000 members (native speakers of Icelandic, 18 years and older) of the net panel from the Social Science Research Institute of the University of Iceland and an online questionnaire was distributed to them via email. 538 individuals answered the online questionnaire in this time period, constituting a response rate of 54%. Prior to the actual survey, listeners were informed that the playschools in the capital area of Reykjavík, Iceland, were looking for employees in a multicultural playschool where employees, as well as children, would be of different nationalities. Respondents were told that three women had been pre-selected as best qualified for the job and that the respondents’ help was required in order to decide between the candidates. Playschools in Iceland are open to children from 18 months until primary school age (5–6 years of age). Over half of the employees working in playschools are unskilled while roughly one third are educated playschool teachers (Statistics Iceland 2017b). 5.7% of all employees working in playschools are men and slightly more than half of employees are between 20–39 years of age (Education 2010: 11), thereby outnumbering other age groups. Thus, the age and gender of the speakers of this study coincide with the most common gender and age groups represented by playschool employees in Iceland. The relatively high personnel turnover in playschools sometimes contributes to the belief that everybody can work in a playschool independent of their educational background. However, this assumption only holds true for playschool personnel with non-administrative duties. 6.3% of all children attending playschools are foreign citizens. In comparison, 6.2% of all employees in Icelandic playschools are foreigners (Statistics Iceland 2017b), matching the percentage of foreign children.

Chapter 3.  Evaluations of foreign-accented Icelandic 59



Iceland has one of the highest birth rates in Europe and has often been characterised as a closely-knit society. For this reason, it can be assumed that every Icelander has a good understanding of how everyday routines in playschools are, either from experience with their own children or children within the family or friends group. Placing the current study within the context of Icelandic playschools, therefore, seemed reasonable both in terms of designing a context that is easily understood by participants and considering that the background of employees in Icelandic playschools fits that of speakers selected for this study. Concerning the execution of the study, each participant listened to three out of the eight audio stimuli with each speaker reading the same text outlined above. The online questionnaire was designed in a way that all participants listened to the native speaker of Icelandic while the other two stimuli were recordings of women non-native speakers of Icelandic. The recordings were played in randomised order. Those participants that claimed to have listened to the whole or most of the recording were able to access and answer questions about it, those that did not make such a claim were disregarded from analysis. In addition, listeners also answered questions about their own personal backgrounds. 4.3

Survey and data analysis

A 7-point semantic differential from 3 to –3 was used, built on the personality traits employed in the MIN project (e.g. Kristiansen 2006). Implementing ANOVA and chi-square analysis, questions presented in this study were tested for significant differences, indicated in the tables as follows: * p ≤ .05, ** p ≤ .01, *** p ≤ .001. Tables also show evaluation means (M) of the semantic differentials for every subgroup, reaching from 3 to –3 according to weighted cases6 and number of participants (N). The adjectives chosen and tested for MIN (Kristiansen 2006: 16; 2010)7 were re-used in the present study. In this survey, however, the eight adjectives on the positive pole were contrasted with appropriate antonyms: ambitious – unambitious, attractive – unattractive, intelligent – stupid, reliable8 – unreliable, efficient – inefficient, independent – dependent, interesting – uninteresting, relaxed – agitated. Two 6. The data was weighted in terms of gender, age, residency and education in order to be able to transfer results from the sample to the population and draw according conclusions that can be generalised for the whole of the Icelandic population. 7. Adjectives for testing were originally Danish and were translated into the other Nordic languages. 8. The original adjective here was confidence-inspiring (Icelandic: traustvekjandi) but it was replaced by another, yet related, adjective (Icelandic: áreiðanlegur) in order to be able to find an appropriate antonym.

60 Stefanie Bade

aspects were decisive in resorting to the adjectives employed in the MIN project: Firstly, attitude research as presented in this study is new to the Icelandic context, which is why it draws on the experience of those who designed and conducted the MIN project. Secondly, this study is incorporated into a series of projects focusing on attitudes towards regional language variation in Iceland and attitudes towards formal and informal language use in social media in Iceland. All three studies employ the same set of adjectives in order to make them comparable.9 5. Results 5.1

Evaluations of personality traits

When examining the total means for each speaker (see Figure 2), it becomes apparent that the US-American speaker is evaluated highest (M = 1.93) and the Lithuanian speaker lowest (M = 1.31). The Danish (M = 1.92) and Icelandic speaker (M = 1.86) are in second and third place. While the German and Thai speaker share the fourth place (M = 1.74), the Polish (M = 1.63) and Philippine speaker (M = 1.56) follow in sixth and seventh place. 2.5 2

1.93

1.92

1.86

1.74

1.74

1.5

1.63

1.56 1.31

1 0.5 0

USA

Denmark Iceland Germany Thailand

Poland Philippines Lithuania

Figure 2.  Means across all traits for each speaker’s country of origin

9. The categorisation of semantic differential scales into superordinate systems and its choice of traits is not rigid but can vary according to context, aim, and extent of the respective study. For this reason, I am currently designing a qualitative study aiming at collecting ideas and adjectives associated with different accents.

Chapter 3.  Evaluations of foreign-accented Icelandic 61



The stimulus with the native Icelander scores highest in two categories (attractive, relaxed), while the US-American guise (intelligent, reliable, interesting) and the Danish guise (ambitious, efficient, independent) score highest in the other categories. The Lithuanian guise scores lowest in seven of eight traits (see Figure 3). 2.5

Independent Efficient

Intelligent Ambitious

Relaxed Interesting

Attractive Reliable

2 1.5 1 0.5 0

USA

Denmark

Iceland

Germany

Thailand

Poland

Philippines Lithuania

Figure 3.  Results for all stimuli and personality traits

On closer investigation the following pattern unfolds: The Icelandic, US-American, and Danish guise generally take turns in occupying the first three positions. The Philippine and Lithuanian guise generally share last places for these evaluations, with the Philippine guise typically occupying seventh and the Lithuanian guise reaching eighth place. The Thai, German, and Polish guise are usually mid-range. For efficient, the Polish guise achieves second place, while the Icelandic guise reaches sixth place only. 5.2

Gender

Significant differences turned out to be very unevenly distributed across listeners’ gender in relation to an individual speaker. While there are relatively many significant differences detectable for the Icelandic, US-American, Lithuanian, and German speaker, the opposite is the case for the Danish, Philippine, Polish, and Thai speaker (see Table 1). As Table 1 shows, women evaluate the stimuli generally higher than men. There are, however, exceptions to the Danish stimulus for the traits relaxed and attractive as well as the Polish guise for the traits relaxed, interesting, attractive, and reliable. However, these differences are not significant.

Stefanie Bade

Table 1.  Gender and significant differences of listeners across personality traits and nationalities

Iceland   1.83***   1.86***   1.92***   1.95***   1.97***   1.70*** 1.75* Men N = 245–250 1.66 1.69 1.78 1.76 1.79 1.53 1.63 Women N = 215–222 2.03 2.06 2.09 2.18 2.17 1.90 1.88 USA Men N = 74–76 Women N = 62–76

 1.95*   2.01*** 2.02* 1.74 1.72 1.84 2.20 2.35 2.23

 2.09*   1.92***   1.78***   1.73*** 1.89 1.68 1.44 1.47 2.32 2.22 2.17 2.02

Reliable

Attractive

Interesting

Relaxed

Ambitious

Intelligent

Gender

Efficient

Personality traits Independent

62

 1.89* 1.76 2.03 1.95 1.86 2.05

Denmark Men N = 65–66 Women N = 49–53

2.00 1.97 2.05

2.18 2.11 2.27

1.94 1.87 2.01

2.12 2.00 2.27

1.87 1.91 1.80

1.76 1.66 1.88

1.62 1.64 1.60

1.88 1.82 1.95

Philippines Men N = 66–67 Women N = 65–68

1.61 1.54 1.69

  1.82*** 1.53 2.09

1.67 1.63 1.71

1.77 1.71 1.84

1.39 1.36 1.42

1.43 1.25 1.61

1.18 1.07 1.30

1.59 1.42 1.76

Lithuania Men N = 74–76 Women N = 55–59

  1.37***   1.66***   1.40***   1.55***   1.19***   1.11***   0.96***   1.24*** 0.93 1.19 0.82 1.01 0.65 0.58 0.41 0.78 1.95 2.27 2.20 2.26 1.89 1.80 1.68 1.84

Poland Men N = 64–66 Women N = 72–77

1.84 1.77 1.90

2.10 1.97 2.21

1.72 1.62 1.81

1.94 1.81 2.05

1.04 1.27 0.84

1.54 1.57 1.52

1.30 1.32 1.29

1.58 1.61 1.56

Thailand Men N = 55–58 Women N = 62–65

1.77 1.55 1.98

 1.95* 1.66 2.20

 1.79* 1.52 2.03

1.96 1.77 2.14

1.47 1.31 1.62

 1.64* 1.35 1.91

1.53 1.29 1.74

  1.83*** 1.48 2.13

Germany Men N = 79–82 Women N = 60–63

5.3

  1.79***   1.89***  1.93*   1.95***   1.76***   1.67***   1.35***   1.57*** 1.47 1.54 1.73 1.60 1.51 1.31 1.02 1.21 2.20 2.33 2.21 2.43 2.11 2.15 1.79 2.04

Age

Outcomes for age show a similar pattern when considering the distribution of significant differences (see Table 2). Generally, the pattern follows that the older the listeners the higher they evaluate the eight personality traits. This pattern is especially valid for the Icelandic, US-American, and Danish guise. There are, however, exceptions. The traits independent, efficient, and ambitious for the Polish guise exhibit a somewhat reverse pattern with the youngest age group evaluating those traits higher than the other age groups. Differences are, however, not significant.

Chapter 3.  Evaluations of foreign-accented Icelandic 63



Table 2.  Gender and significant differences across personality traits and nationalities

Reliable

Attractive

Interesting

Relaxed

Ambitious

Intelligent

Efficient

Age

Independent

Personality traits

Iceland 18–29 N = 104–105 30–44 N = 134–136 45–59 N = 107–113 60+ N = 113–121

  1.83***   1.86***   1.92***   1.95***   1.97*   1.70***   1.75***   1.89*** 1.66 1.68 1.75 1.69 1.82 1.40 1.49 1.69

USA 18–29 N = 34 30–44 N = 33 45–59 N = 29–32 60+ N = 41–43

  1.95***   2.01*** 1.54 1.52

1.56

1.65

1.77

1.71

1.79

1.51

1.58

1.64

1.96

1.98

1.98

2.15

2.12

1.78

1.82

2.06

2.19

2.16

2.21

2.29

2.15

2.13

2.10

2.17

2.02 1.78

2.09 1.82

  1.92***   1.78***   1.73*** 1.67 1.11 1.15

 1.95* 1.81

1.68

1.70

1.84

1.87

1.47

1.56

1.47

1.49

2.10

2.31

2.19

2.40

2.19

2.00

2.11

2.10

2.38

2.42

2.24

2.23

2.29

2.31

2.10

2.32

Denmark 18–29 N = 24 30–44 N = 38 45–59 N = 28 60+ N = 24–29

2.00 1.75

2.18 1.86

1.94 1.49

2.12 1.75

1.87 1.38

 1.76* 1.02

 1.62* 0.91

  1.88*** 1.25

1.96

2.04

1.98

2.19

1.91

1.79

1.68

1.77

2.19

2.39

1.98

2.25

2.01

2.08

1.85

2.14

2.10

2.46

2.23

2.21

2.12

2.06

2.00

2.29

Philippines 18–29 N = 23–25 30–44 N = 48 45–59 N = 28 60+ N = 32–35

   1.61* 0.98

1.82 1.35

1.67 1.20

1.77 1.50

1.39 1.15

 1.43* 1.01

 1.18* 0.55

 1.59* 1.10

1.57

1.78

1.63

1.70

1.23

1.16

1.11

1.42

1.95

2.15

2.03

2.04

1.42

1.82

1.23

1.92

1.85

1.91

1.76

1.86

1.77

1.77

1.72

1.88

(continued)

64 Stefanie Bade

Table 2.  (continued)

Reliable

Attractive

Interesting

Relaxed

Ambitious

Intelligent

Efficient

Age

Independent

Personality traits

  1.37***  1.66*   1.40*** 1.42 1.61 1.14

1.55 1.37

0.69

1.22

0.82

1.18

0.36

0.28

–0.07

0.55

1.61

1.93

1.88

1.93

1.49

1.48

1.42

1.84

1.93

2.01

2.01

1.86

1.98

1.90

1.98

1.81

Poland 18–29 N = 33–34 30–44 N = 34 45–59 N = 37–41 60+ N = 33–35

1.84 2.21

2.10 2.23

1.72 1.73

1.94 2.23

1.04 1.02

1.54 1.35

1.30 0.99

1.58 1.39

1.69

2.01

1.63

1.77

1.01

1.53

1.31

1.44

1.87

2.02

1.76

1.95

1.09

1.78

1.43

1.80

1.59

2.15

1.77

1.81

1.05

1.47

1.46

1.68

Thailand 18–29 N = 20 30–44 N = 29–30 45–59 N = 35–38 60+ N = 34–36

1.77 1.42

1.95 1.83

1.79 1.64

 1.96* 1.88

1.47 1.31

1.85

2.20

2.00

2.38

1.22

1.67

1.65

1.93

1.62

1.75

1.45

1.47

1.35

1.39

0.95

1.59

2.08

2.02

2.03

2.17

1.90

2.12

2.10

2.14

Germany 18–29 N = 34 30–44 N = 44–45 45–59 N = 29–33 60+ N = 32–35

1.79 1.68

Lithuania 18–29 N = 36 30–44 N = 38 45–59 N = 21–23 60+ N = 34–37

  1.19***   1.11***   0.96***   1.24*** 1.08 0.95 0.75 1.05

 1.64*   1.53*** 1.19 1.40

1.83 1.56

  1.89***  1.93*   1.95***   1.76***   1.67***   1.35***   1.57*** 1.48 2.08 1.53 1.67 1.81 0.84 0.95

1.49

1.55

1.51

1.57

1.34

1.12

1.01

1.38

2.12

2.41

2.23

2.50

2.05

1.95

1.63

2.10

2.02

2.24

2.10

2.39

2.19

2.03

2.08

1.96



Chapter 3.  Evaluations of foreign-accented Icelandic 65

6. Discussion Several conclusions can be drawn from the analysis of the background variables gender and age. Firstly, since the native Icelandic guise is rated lower than some non-native guises a strict distinction built on nativeness/non-nativeness alone does not seem to be appropriate. Secondly, the evaluations of certain accents showed greater variability than others. Remarkably enough, the native Icelandic accent belongs to the more variable group. This great variability among the Icelandic, US-American, German and Lithuanian guise, which can be observed for both background variables investigated in this study, is quite puzzling and requires further investigation. Thirdly, the variability of the evaluation of some accents shows that the listeners have varying attitudes to some accents rather than to others. Therefore, we must assume that individual accents conjure up different images or stereotypes connected with speakers’ nationality. While it was suggested earlier that women tend to prefer standard forms over men, the opposite seems to be true for this study. In this context, however, we need to take into consideration that since all speakers in this study were women, we may be seeing a sympathy effect on the part of female listeners. Similarly, it is conceivable that Icelandic men do not relate to the work as a playschool teacher as much as Icelandic women do, considering that only about every twentieth playschool employee is male (see 4.2). The profession of a playschool teacher could be connected with female dominance, resulting in men’s general downgrading of the voices. This would constitute an opposite effect to the sympathy effect on the part of female listeners outlined above. It would, therefore, be advisable to conduct a similar study with male speakers or in a different contextual setting. Regarding the latter aspect, it is quite possible that participants deemed native Icelanders, as represented by a native Icelandic accent, inappropriate for employment in a multicultural playschool, thereby explaining the downrating of the guise. Another methodological aspect that could be considered in future studies is an alternative approach to the analysis of participants’ background. While an earlier study on attitudes towards influences from English employed a lifestyle model (cf. Óladóttir 2009: 33ff.) which did not seem fully appropriate for the Icelandic context, the concept of cosmopolitan vs. localist (cf. Stefánsdóttir 2018)10 seems promising with regard to multiculturalism and language variation.

10. This concept is currently being used in the ongoing project Modelling the Linguistic Consequences of Digital Language Contact in the context of cultural motivation, i.e. how an individual’s identity in cultural contexts influences attitudes towards Icelandic and English.

66 Stefanie Bade

When examining the four age groups, there is a noticeable distinction between the two younger groups (18–44) evaluating both native and non-native speakers generally lower than the other two age groups (45+). These results are surprising in that a reverse relationship was expected as indicated by former findings (cf. Bade and Isenmann unpublished; Kristiansen 2006), especially in terms of the younger age group evaluating non-standard more positively than older age groups. An exception to this is found in the evaluations of the Polish guise concerning the traits independent, efficient and ambitious, all of them representing attributes connected with competence, which score highest among the youngest age group. Considering the ratings for all traits across countries the Polish guise scores second highest for efficiency, a result that might be traceable to Polish immigrants often being portrayed as effective and hard-working in Icelandic media. Hence, it is conceivable that listening to a Polish accent triggers a positive stereotype about Polish immigrants to Iceland. One-half (51%) correctly identified this speaker as a native-speaker of Polish which is the second highest score for correct identification apart from the native-speaker of Icelandic (90%). As already indicated, background of speakers was hidden to participants in this study. Scores for correct identification of other speakers’ country of origin were generally very low ranging from 2% and 3% for the Thai and US-American guise to 7% and 8% for the Lithuanian and Danish guise. 27% of all participants correctly identified the Philippine and German guise. A correlation analysis between correctly identified countries of origin and evaluation of accents is yet to be carried out. Returning to the results for the other guises, it is imperative to bear in mind that Icelandic society and its speech community has undergone enormous demographic changes that have transformed it from a traditionally monoethnic and monolingual society to one that is multiethnic and multilingual. It is therefore conceivable that increased competition in the labour market and higher job requirements cause natives to experience insecurity from competing, non-native individuals that is reflected in evaluations of non-native speakers. In that respect, middle age groups have been said to bear most responsibility within Icelandic society (cf. Ewen and Kristiansen 2006: 47), which could be a potential further cause for social insecurity among those and even among the youngest age groups as a consequence of societal changes. It is, however, striking that younger age groups are more negative towards all guises, irrespective of the nativeness/non-nativeness of the accent. Could the social insecurity described above include the native Icelandic accent? Bearing in mind that the native Icelandic accent caused much disagreement in evaluation among participants and that it was rated lower than the US-American and Danish accents, when considering means across all traits, the question arises whether a strict native/non-native distinction may not be at stake here, but instead puristic concepts considered preeminent for the linguistic landscape in Iceland, are in fact



Chapter 3.  Evaluations of foreign-accented Icelandic 67

being challenged as well as the status of the Icelandic language in general. Is there a divergence between positive overt attitudes towards linguistic purism and covert disagreement with it that can be traced to changed circumstances, linguistically, socially and culturally? When considering both the means for each speaker’s country of origin across all traits and those for gender and age, it is striking that what we could subsume under the category Western accents seems to be generally preferred over others, i.e. Eastern European and Asian accents. A similar result was found by Lindemann (2005). Additionally, the strict native/non-native distinction suggested by Moyer (2013: 102) does not seem to be applicable to the Icelandic context, since the Icelandic guise does not perform in a way that would justify a strict distinction, but can rather be located within a broader continuum of perceived in-group identity (cf. Social Identity Theory, Tajfel and Turner 1979). Considering historical as well as current sociopolitical and cultural circumstances, the high ratings for the US-American and Danish guises do not seem to be that far-fetched. Although overt attitudes towards Danish and the instruction of Danish in secondary education are fairly negative among younger people,11 covert attitudes towards Danish paint a different picture (Ragnarsdóttir and Oddsdóttir 2016). As for the high rating of the US-American guise, reasons might be found in both political-historical circumstances and the USA’s current role as cultural leader and role model. Since the US-American occupation of Iceland during World War II, Icelandic daily life has increasingly been influenced by US-American culture with consequences for the Icelandic language. US-American material makes up for almost half of the broadcasting time on Icelandic television resulting in substantial exposure of Icelandic viewers to US-American English from an early age. Along with other factors such as high usage of social media platforms (Kristinsson 2013) as well as using English to a large extent in certain domains (Arnbjörnsdóttir and Ingvarsdóttir 2018), English does not only exercise an influence on corpus-related elements of the Icelandic language, but exerts pressure on status-related aspects, implying an increasing rise in prestige for the English language. To sum up, a categorisation of perceived accent using the concept of nativeness/ non-nativeness is only conceivable when taking factors such as familiarity, common historical and cultural bonds as well as prestige into account. On the whole, these elements (presumably along with others) seem to form the basis of an in-group/ out-group distinction that resembles ideas about one’s own (nativeness) and the other (non-nativeness) reflected in and triggered by accent. 11. Danish is taught as a second mandatory foreign language in Icelandic schools in order to perpetuate and promote relations with the other Nordic countries on the basis of a common culture and history (cf. Ministry of Education 2007: 29).

68 Stefanie Bade

7. Conclusion Drastic demographic changes have transformed Icelandic society from a monolingual to a multilingual society with accented speech becoming more and more apparent in everyday life. Evaluations of Icelandic that were characterised by a standard/non-standard dichotomy for centuries might make way for a different, more complex model based on an affinity for speakers in terms of perceived degree of nativeness/non-nativeness built on ideas about in-group/out-group. Thus some non-native accents could be stigmatised while others enjoy high prestige, even comparable to or exceeding that of the native variety.

References Adamson, Hugh Douglas and Vera Regan. 1991. The acquisition of community speech norms by Asian immigrants learning English as a second language: A preliminary study. Studies in Second Language Acquisition 13. 1–22.  https://doi.org/10.1017/S0272263100009694 Arnbjörnsdóttir, Birna and Hafdís Ingvarsdóttir. 2018. Language development across the lifespan: The impact of English on education and work in Iceland. Educational Linguistics 34. Cham: Springer. Árnason, Kristján. 2003. Icelandic. In Ana Deumert and Wim Vandenbussche (eds.), Germanic standardizations: Past to present, 245–279. Amsterdam-Philadelphia: John Benjamins Pub­ lishing.  https://doi.org/10.1075/impact.18.10arn Árnason, Kristján. 2005. Íslensk tunga 1, hljóð: Handbók um hljóðfræði og hljóðkerfisfræði [Icelandic language 1, sounds. Handbook of phonology and phonetics]. Reykjavík: Almenna Bó­ka­félagið. Árnason Kristján. 2006. Island. In Tore Kristiansen and Lars S. Vikør (eds.), Nordiske Språk­hald­ ningar: Ei meiningsmåling, 17–39. Oslo: Novus. Bade, Stefanie and Vanessa Isenmann. Good and not so good Icelandic. Standard Icelandic and evaluations of linguistic variation with focus on foreign-accented speech and computermediated communication. Unpublished manuscript, 2.12.2016. Brennan, Eileen M. and John S. Brennan. 1981. Accent scaling and language attitudes: Reactions to Mexican American English speech. Language and Speech 24 (3). 207–221. https://doi.org/10.1177/002383098102400301 Cargile, Aaron and Howard Giles. 1998. Language attitudes toward varieties of English: An American-Japanese context. Journal of Applied Communication Research 26. 338–356. https://doi.org/10.1080/00909889809365511 Education [Skólamál]. 2010. Starfsfólk í leikskólum 2009. [Personnel in pre-primary schools.] Hagtíðindi [Statistical Series] 2. 1–16. Ewen, Halldóra Björt and Tore Kristiansen. 2006. Island. In Tore Kristiansen (ed.), Nordiske sprogholdninger: En masketest, 33–48. Oslo: Novus. Flege, James E., Murray J. Munro and Ian R. A. MacKay. 1995. Factors affecting strength of perceived accent in a second language. Journal of the Acoustical Society of America 97 (5). 3125–3134.  https://doi.org/10.1121/1.413041



Chapter 3.  Evaluations of foreign-accented Icelandic 69

Garrett, Peter. 2010. Attitudes to language. Cambridge: Cambridge University Press. https://doi.org/10.1017/CBO9780511844713 Giles, Howard and Nancy Niedzielski. 1998. Italian is beautiful, German is ugly. In Laurie Bauer and Peter Trudgill (eds.), Language myths, 85–93. London: Penguin. Hilmarsson-Dunn, Amanda and Ari Páll Kristinsson. 2013. The language situation in Iceland. In Robert M. Kaplan, Richard B. Baldauf , Jr. and Nkonko M. Kamwangamalu (eds.), Language planning in Europe, 100–169. London-New York: Routledge. Hlynsdóttir, Kristín Ingibjörg. 2016. Þykir harðmæli betra en linmæli? Rannsókn á ómeðvitum viðhorfum. [Is hard speech better than soft speech? An investigation of covert attitudes.] (Bachelor Thesis in Icelandic Linguistics. University of Iceland). Retrieved from http://hdl. handle.net/1946/26063. Kang, Okim and Donald Rubin. 2009. Reverse linguistic stereotyping: Measuring the effect of listener expectations on speech evaluation. Journal of Language and Social Psychology 28 (4). 441–456.  https://doi.org/10.1177/0261927X09341950 Kristiansen, Tore (ed.) 2006: Nordiske sprogholdninger: En masketest. Oslo: Novus. Kristiansen, Tore and Lars S. Vikør. 2006. Nordiske språkhaldningar – jamføring og konklusjonar”. In Tore Kristiansen and Lars S. Vikør (eds.), Nordiske Språkhaldningar: Ei meiningsmåling, 199–214. Oslo: Novus. Kristínsson, Ari Páll. 2013. Evolving language ideologies and media practices in Iceland. Sociolinguistica 27. 54–68. Ladegaard, Hans J. 1998. National stereotypes and language attitudes: The perception of British, American and Australian language and culture in Denmark. Language and Communication 18. 251–274.  https://doi.org/10.1016/S0271-5309(98)00008-1 Lambert, Wallace E., Richard C. Hodgson, Robert C. Gardner and Samuel Fillenbaum. 1960. Evaluational reactions to spoken languages. Journal of Abnormal and Social Psychology 60 (1). 44–51.  https://doi.org/10.1037/h0044430 Leonard, Pax Stephen and Kristján Árnason. 2012. Language ideology and standardisation in Iceland. In Tore Kristiansen and Nikolas Coupland (eds.), Standard languages and language standards in a changing Europe, 91–96. Oslo: Novus. Lindemann, Stephanie. 2005. Who speaks ‘broken English’? US undergraduates’ perceptions of non-native English. International Journal of Applied Linguistics 15 (2). 187–212. https://doi.org/10.1111/j.1473-4192.2005.00087.x Ministry of Education [Menntamálaráðuneyti]. 2007. Aðalnámskrá grunnskóla: Erlend tungumál. [National Curriculum for primary schools: Foreign languages.] Moyer, Alene. 2013. Foreign accent. The phenomenon of foreign-accented speech. Cambridge: Cambridge University Press.  https://doi.org/10.1017/CBO9780511794407 Óladóttir, Hanna. 2009. Shake, sjeik eller mjólkurhristingur? Islandske holdninger til engelsk språkpåvirkning. Oslo: Novus. Purcell, Edward T. and Richard W. Suter. 1980. Predictors of pronunciation accuracy: A reexamination. Language Learning 30. 271–287.  https://doi.org/10.1111/j.1467-1770.1980.tb00319.x Ragnarsdóttir, Brynhildur Anna and Þórhildur Oddsdóttir. 2016. Afstaða nemenda í 10. Bekk til Norðurlandamála. [Attitudes of 10th grade students towards Nordic languages.] Málfríður 32 (2). 18–20. Ryan, Ellen B. 1983. Social psychological mechanisms underlying native speaker evaluations of non-native speech. Studies in Second Language Acquisition 5 (2). 148–159. https://doi.org/10.1017/S0272263100004824

70 Stefanie Bade

Ryan, Ellen B. and Miguel Carranza. 1975. Evaluative reactions of adolescents toward speakers of standard English and Mexican accented English. Journal of Personality and Social Psychology 31 (5). 855–863.  https://doi.org/10.1037/h0076704 Ryan, Ellen B. and Richard J. Sebastian. 1980. The effects of speech style and social class background on social judgments of speakers. British Journal of Social and Clinical Psychology 19. 229–233.  https://doi.org/10.1111/j.2044-8260.1980.tb00348.x Statistics Iceland. 2017a. Mannfjöldi eftir fæðingarlandi. [Population according to the country of birth.] Retrieved from http://www.hagstofa.is, accessed October 13, 2017. Statistics Iceland. 2017b. Leikskólabörnum og starfsfólki fækkar á leikskólum fækkar. [Number of children and employees in playschools decreases.] Retrieved from https://hagstofa.is/utgafur/ frettasafn/menntun/leikskolar–2016/, accessed May 3, 2018. Stefánsdóttir, Lilja Björk. 2018. Heimdragar og heimsborgarar. Menningarlegur hvati í stafrænu málsambýli. [Localists and cosmopolitans. Cultural motivation in digital language contact.]. (Master Thesis in Icelandic Linguistics. University of Iceland, unpublished). Tajfel, Henri and John Turner. 1979. The social identity theory of intergroup behaviour. In William G. Austin and Stephen Worchel(eds.), The social psychology of intergroup relations, 7–24. Chicago: Nelson Hall. White, Michael J. and Yan Li. 1991. Second-language fluency and person perception in China and the United States. Journal of Language and Social Psychology 10 (2). 99–113. https://doi.org/10.1177/0261927X91102002

Chapter 4

C’era i fascisti e i tedeschi Instances of linguistic simplification in a corpus of Italiano popolare Silvia Ballarè and Eugenio Goria1 University of Torino

The aim of this paper is the discussion of preliminary results of the ongoing study of an Italian substandard variety (italiano popolare) with the ParVa spoken corpus (Guerini 2016). The focus is on the interplay of two structural factors: contact with Italo-Romance dialects and linguistic simplification. The corpus shows previously overlooked syntactic features, which may improve the view of this subvariety and also allows a quantitative analysis of agreement in existential constructions. Keywords: italiano popolare, social variation, simplification, existential constructions

1. Introduction Over the past century, the Italian lowermost variety on the diastratic axis of variation (italiano popolare) has been closely studied. This varietal development has occurred mainly by the relatively recent schooling of a large part of the population and the survival of the regional Italo-Romance (primary) dialects alongside the standard variety. These dialects still represent the L1 for lower-class speakers in several regions. Due to this specific sociolinguistic configuration, italiano popolare has over time developed a number of diagnostic features, some of which are the result of substrate influence from a regional dialect, while others represent the outcome of a more general process, namely linguistic simplification. However, until recently research on italiano popolare relied primarily on written data. Only recently spoken

1. This paper is the result of ongoing and systematic collaboration between its two authors. Silvia Ballarè wrote Sections 0, 1 and 4, while Eugenio Goria wrote Sections 2, 3 and 5. https://doi.org/10.1075/silv.22.04bal © 2019 John Benjamins Publishing Company

72

Silvia Ballarè and Eugenio Goria

corpora became available, making it possible to test previous assumptions and to discover previously unaccounted features. In this paper, our use of the Parva corpus (Partigiani Val Camonica, Guerini 2016) aims at both goals. This corpus has about 15 hours of semi-structured interviews on the Resistance during World War II in Northern Italy (particularly in the region of Val Camonica, province of Brescia). The interviewees are 24 former resistance fighters of both genders who share their war memories. Most relevant to our research is that they provide extensive biographical information, particularly about their jobs and education, thus allowing the identification of class distinctions. In Section 2 and 3 we discuss the major features of italiano popolare, paying special attention to the phenomenon of linguistic simplification. In Section 4 we outline our research questions and provide a description of the methodology used for the analysis. Section 5 first presents the outcomes of a qualitative investigation involving this corpus (Section 5.1), followed by a case study concerning a single linguistic variable (Section 5.2). 2. On italiano popolare In Italian sociolinguistics, italiano popolare refers to a social variety associated with speakers with a limited educational background. This variety received a lot of attention during the second half of the 20th century (see e.g. De Mauro 1970[1963]; Cortelazzo 1972; Sanga 1980 and Berruto 1983), and was dealt with mainly in relation to speakers with a diglossic linguistic repertoire. Typically, such speakers have an Italian dialect as their L1 and acquired Standard Italian only as a second language and through the educational system. It is important to stress that Italo-Romance dialects are not varieties of Italian but, according to Coseriu (1980), they are to be considered as primary dialects, i.e. languages which are distinct from Italian, given their significant structural differences (see Manzini and Savoia 2005). The most widespread type of linguistic repertoire on the Italian peninsula has a dilalic structure (see dilalia in Berruto 1987), with both Italian and one or more dialects being equally present in early socialisation and informal conversation. Therefore, it can be observed that while in a broader perspective italiano popolare is clearly diastratically marked, in that it is associated with speakers with limited formal education, if we focus on this specific subgroup of the speech community, the picture that emerges is different. In fact, for these speakers italiano popolare represents the H variety, given that it is the only variety of Italian that they master. Since it is used only in formal situations and represents the highest degree of formality that these speakers can achieve, it also displays strong diaphasic markedness.

Chapter 4.  C’era i fascisti e i tedeschi 73

Due to the structure of the speakers’ repertoire (Cerruti and Regis 2014) and to the fact that in the Italian context social variation is subordinate to geographic variation (see Koch and Oesterreicher 1990 and Berruto 2012 [1987]), it would be opportune to consider several different regional varieties of italiano popolare2 in relation to the dialectological background of specific areas. In other words, every variety of italiano popolare represents a regional variety and is therefore marked diatopically. In fact, substrate influence from Italian dialects is one of the main sources responsible for the emergence of substandard features in this variety. At the same time, however, all the varieties of italiano popolare also display great structural similarities, irrespective of their substrate dialects. An example of a pan-Italian feature of italiano popolare is the case of doubly filled complementisers, as in (1):3 (1) a. ItaPop:     quando che è venuta la liberazione when comp be:3sg come.pp the liberation b. StIta:     quando è venuta la liberazione when be:3sg come.pp the liberation ‘When the Liberation came’

This linguistic feature is discussed by Cerruti (2011: 15), who states that “due to its over-regional presence, it can be considered as one of those fossilized interferences that form the basis for the […] common grammar of italiano regionale popolare.” In several cases the features that have emerged independently from substrate influence are the result of a process of linguistic simplification (Berruto 1983: 15), whereby they contain alternatives which are structurally lighter or less complex than those occurring in the standard variety of Italian. Such a process will be dealt with in more detail in the following section.

2. Each geographical variety has its own internal structure dictated by the crossing of the diastratic and the diaphasic axes. However, as far as italiano popolare is concerned, diatopic factors bear much greater weight than they do in other varieties. 3. Throughout the paper, the Italian examples will have the same standardized format. Interlinear glosses with respect to the Leipzig glossing rules (https://www.eva.mpg.de/lingua/pdf/ Glossing-Rules.pdf), and a list of abbreviations is provided at the end of the paper. The examples labelled as ItaPop come from the working-class subcorpus of the ParVa corpus (see below) and have not been edited for publication, while their Standard Italian counterparts are provided by the authors.

74

Silvia Ballarè and Eugenio Goria

3. Linguistic simplification in italiano popolare As previously stated, italiano popolare differs from other sociolects of Italian, first of all because of the greater influence of substrate dialects, and secondly because of linguistic simplification. These are also the main points whereby it differs structurally from Standard Italian, which is by definition supra-regional and structurally elaborate. It is important to note that the effects of language contact are highly unpredictable, since the extent of dialect influence varies according to the structure of regional (or perhaps sub-regional) linguistic repertoires. Moreover, contact with other Italo-Romance varieties produces different results in different situations, including matter and pattern replication (Matras and Sakel 2007), as well as contact-induced grammaticalisation (Cerruti 2014). Conversely, if a typologically-oriented notion of linguistic complexity is adopted, as in McWorther (2001) and Miestamo (2008, 2017), it is possible to disentangle the outcomes of linguistic simplification from other phenomena occurring in sub-standard varieties. Within this framework, our focus will be on what Miestamo (2017: 239) terms local complexity. Therefore, we will identify formal parameters that allow us to consider the grammatical encoding of one particular functional domain as simpler or more complex than other grammatical strategies, as opposed to adopting holistic complexity metrics, e.g. in Kortmann and Szmrecsanyi (2009), to demonstrate that italiano popolare as a variety is structurally simpler than other varieties of Italian. Our claim is much more limited in this respect, for we argue that the constructions involved in the marking of single functional domains in italiano popolare represent structurally simpler alternatives than those occurring in the standard variety. On the contrary, extending this type of reasoning to the whole variety would involve a number of theoretical problems. Namely, within the Italian sociolinguistic scenario, substrate influence of the local dialects constantly interacts with contact-independent simplification, and the combination of the two could actually result in both simplification and complexification with respect to the standard. The parameters on which we rely for our account of simplification are the two diagnostic features of complexity discussed by Miestamo (2008), which can be paraphrased as follows: a. “Fewer distinctions”: the more distinctions are grammaticalised within a functional domain, the greater the complexity; b. “One meaning one form”: multiple encoding of the same function is more complex than single encoding Furthermore, while these are generally intended paradigmatically, in terms of a loss of structural oppositions, allomorphy, etc., we argue that simplification may

Chapter 4.  C’era i fascisti e i tedeschi 75

also be observed at the syntagmatic level. This involves parameter (ii) in particular, e.g. those cases in which single marking of a grammatical category is preferred to double marking. A good example of simplification on the syntagmatic level is the development of post-verbal negation in contexts where Standard Italian only allows discontinuous forms4 (see Sanga 1980: 6 and Cortelazzo 1972: 107). This provides a structurally simpler alternative, in that single-marking of negation is preferred to double-marking. Consider, for example, the difference between the structures in (2a) and (2b): (2) a. StIta:     non mi sono fatto niente neg 1sg.obl be1sg do:pp nothing b. ItaPop:     mi sono fatto niente 1sg.obl be:1sg do:pp nothing ‘I did not hurt myself ’

Simplification at the paradigmatic level can be found in the system of relative pronouns: while Standard Italian requires inflected relative pronouns, as in (2a), the marking of relative clauses is done in italiano popolare by means of uninflected general subordinator che, as in (3b); see Alisova (1965): (3) a. StIta:     Cesare a cui è venuta l’idea Cesare to rel.obl be:3sg come:pp the idea b. ItaPop:     Cesare che è venuta l’idea Cesare comp be:3sg come:pp the idea ‘Cesare, who had the idea’

In brief, adopting a typological perspective on linguistic simplification may shed light on the dynamics that shape the structure of italiano popolare. At the same time, one should bear in mind that forces other than linguistic simplification have an influence on this variety, including contact with local dialects and performance-related phenomena. Therefore, linguistic typology can provide an 4. According to Bernini and Ramat (1996: 41), there are pragmatic factors that may favour the occurrence of postverbal negation in languages like Italian, where standard negation can be expressed by a discontinuous strategy. For an account of a similar phenomenon in other varieties of Italian see also Ballarè (2015).

76

Silvia Ballarè and Eugenio Goria

explanation for why italiano popolare is characterised so saliently by particular structures and possibly why sub-standard varieties of different languages share structural similarities. However, other competing motivations will at some future point have to be considered in order to provide an extensive account of this variety. 4. Research questions and methodology In this paper, we will focus on linguistic simplification as one of the mechanisms responsible for the emergence of substandard features in italiano popolare. Particularly, we aim to determine whether newer oral corpora can provide a broader picture of this variety, revealing previously overlooked features. In fact, research in the past decades has been strongly biased towards written language, mostly due to the fact that the vast majority of spoken Italian corpora do not provide any metadata for the speakers (see Crocco 2015 and Goria and Mauri 2018 for a recent overview). On the contrary, written production from speakers of italiano popolare has been widely investigated, e.g. by Sanga (1980). The corpus used for this research is the ParVa corpus (Guerini 2016). This resource tool includes about 15 hours of taped interviews collected in one particular region, namely the Val Camonica (province of Brescia), and involving former fighters who took part in the Resistance during World War II. While there is no regional variation between the speakers, who all belong to the same area, the corpus is particularly fit for investigating the diastratic dimension, in that some of the speakers belong to the lower social classes and have lower educational achievements, while some other are middle-class speakers, as demonstrated by professional qualifications or university degrees. Therefore, as in Cerruti (2016), we have created two subcorpora based on the level of education, the first representing speakers belonging to lower social classes (14 informants), and the second representing middle-class speakers, who have attained a high level of education (10 informants). The analysis presented here is both qualitative and quantitative in scope. In qualitative terms, a manual examination of the corpus yielded an updated inventory of features of italiano popolare produced by structural simplification. Some features are already known from prior descriptions of this variety, while others were previously unaccounted. For the quantitative portion of our study, we provide a thorough examination of a single linguistic variable, namely subject-verb agreement in presentative constructions.

Chapter 4.  C’era i fascisti e i tedeschi 77

5. Analysis 5.1

New features

A manual examination of the working-class component of the ParVa corpus offered a picture that was fairly consistent with previous accounts of italiano popolare. Due to space limitations, our focus will be on two features that have heretofore not been described in relation to the variety in question, and for which we hypothesise a direct connection with speakers of italiano popolare, namely the loss of the reflexive in pronominal verbs and the loss of auxiliaries in the analytic past tense. As with several other Romance languages, Italian has reflexive pronominal verbs that are constructed with a pronominal clitic added to the verbal form. However, such forms may be simplified through the elimination of the pronoun, as shown in the comparison between (4a) and (4b): (4) a. ItaPop:     come siamo andati su, siamo addormentati dopo as be:1pl go:pp up be:1pl fall_asleep:pp later b. StIta:     ci=siamo addormentati dopo 1pl.refl.be:1pl fall_asleep:pp later ‘As we went up, we fell asleep later’

It is important to note that in the example the omission of the reflexive pronoun does not leave room for any other interpretation; the fact that the verb addormentarsi exhibits a reflexive morphology is to be considered an idiomatic characteristic of the verb, in that its semantics does not convey any reflexive meaning. Furthermore, its transitive counterpart (addormentare, ‘put someone to sleep’) requires a different paradigm, i.e. the auxiliary to have (abbiamo addormentato) and not to be. A similar case is represented by the loss of auxiliary verbs in analytic past tense constructions. Standard Italian has a past form characterised by either have + past participle or be + past participle. This subsection of the ParVa corpus contains several examples whose construction is simplified through the omission of the auxiliary form, as in (5): (5) a. ItaPop:     il giorno dopo cominciato a pensare the day after start:pp to think:inf b. StIta:     il giorno dopo ho cominciato a pensare the dayafter have:pres.1sg start:pp to think:inf ‘the day after I started to think’

78

Silvia Ballarè and Eugenio Goria

Both features can be regarded as instances of simplification on the morphosyntactic level. In the first case, the elimination of reflexive clitics may be seen as an elimination of redundancy in person marking, as well as an increase in the semantic transparency of the construction. Furthermore, the same value is conveyed by the auxiliary to be and by its morphology. Similarly, the elimination of the past tense auxiliary in (5) may be considered in relation to the fewer distinctions principle, in which a single form, unmarked for person, is generalised over the whole past-tense paradigm. On a syntagmatic level, only the lexical part of the verb (the past participle) is maintained, while the grammatical morpheme (the auxiliary) is omitted. Finally, it should also be noted that in a case like (5), pragmatics plays a key role; in fact, even if the person marker is lost, the grammatical subject can easily be inferred by the co(n)text, given that it corresponds to the speaker. Such substandard linguistic features only occur in the working-class section of the corpus, and in the literature, they have gone unaccounted in other varieties of Italian. Clearly, the correlation between these features and italiano popolare still has to be demonstrated through quantitative studies testifying that such phenomena are associated with particular groups of speakers or particular sociolects in a statistically relevant way. 5.2

Case study: Existential constructions

Existential constructions (Bentley et al. 2013) have the function of predicating the existence of a particular referent or group of referents. In Standard Italian, this construction has the form of a copular clause, in which the copula is preceded by the fixed clitic pronoun ci, originally of locative value. The referent whose existence is predicated normally occurs after the copula. It is crucial in Standard Italian that there be number agreement between the copula and the following NP, as in (6): (6) ci sono dei libr-i sul tavolo there be: pres.3pl some book-pl on_the table ‘There are some books on the table’  (Bentley et al. 2013: 1. Gloss adapted from the original)

The absence of number agreement between the subject and the verb in existential constructions can be considered a product of linguistic simplification, in that the singular form is systematically overextended to plural controllers and is treated as a default choice, as in (7): (7) ItaPop:   e poi c’è le munizion-i and then there be:pres.3sg the ammunition-pl ‘And then there is the ammunition’

Chapter 4.  C’era i fascisti e i tedeschi 79

In our analysis, we consider agreement in the existential construction to be a sociolinguistic variable, whose variants are represented by the presence or absence of agreement with plural controllers. Previous studies, such as Berruto and Cerruti (2015), have identified the absence of agreement in this context as a feature of several substandard varieties of Italian. The same sociolinguistic markedness is also found in substandard varieties of English (see e.g. Tagliamonte 2009; Tagliamonte and Baayen 2012 and Adger and Smith 2010) and has been considered in relation to “vernacular universals” (Chambers 2004). It must be added, though, that within the Italian context the absence of agreement in this context could also represent a contact feature arising from the local dialects, as is the case in Val Camonica; in fact, in the dialect of this area the existential construction does not exhibit number agreement. Nevertheless, since the phenomenon is not limited to such areas but instead has a much broader distribution, both intra- and extra-linguistically, considering language contact as the only explanation does not tell the whole story. In order to perform a thorough corpus analysis, we have selected the following six factors, both linguistic (following Berruto and Cerruti 2015) and extra-linguistic: Structural factors: – Noun phrase structure: – No determiner (c’è Maria, ‘there is Maria’) – 1 determiner (c’è la ragazza, ‘there is the girl’) – > 1 determiner (c’è la bella ragazza, ‘there is the beautiful girl’) – Coordinated noun phrase (ci sono Maria e Giulia, ‘there is Maria and Giulia’) – Proximity between the copula and the NP: – adjacency (c’è la ragazza, ‘there is the girl’) – non-adjacency5 (c’è veramente la ragazza, lit. ‘there is really the girl’) – Proximity between the copula and the head of the NP: – adjacency (c’è Maria, ‘there is Maria’) – non-adjacency (c’è la bella Maria, ‘there is the beautiful Maria’) – Verb tense: – past (c’era, ‘there was’) – present (c’è, ‘there is’) – future (ci sarà, ‘there will be’)

5. It must be noted that verb-particle constructions are often considered to be geographically marked and are more frequent in the northern varieties (Cini 2008).

80 Silvia Ballarè and Eugenio Goria

Social factors: – Social class: – middle class – lower class – Gender of the speaker: – female – male The analysis was conducted adopting a chi-squared test in order to verify the correlation between substandard variants and lower-class speakers (external predictors), and then to identify positive correlations with specific linguistic contexts (internal predictors). Finally, we compared our results against those obtained by Berruto and Cerruti (2015), who, however, used a different dataset in their analysis of the same variable. Due to the small size of the corpus, we limited ourselves to measuring the significance of each factor through a chi-squared test (p-value 1 determiner coordinated NP Total

Standard

Sub-standard

Total

Sub-st/St

 24 142  10   5 181

 4 25  4  12  45

 28 167  14  17 226

0.17 0.18 0.4  2.4   

Chapter 4.  C’era i fascisti e i tedeschi 81

Even if, as stated above, it is impossible to make an absolute statement based on such a limited data set, it is nevertheless worth noting that the Sub-standard/Standard ratio is significantly higher in the case of coordinated NPs, as in Example (8): (8) ItaPop:   c’era i fascist-i e i tedesch-i there be:pst.3pl the fascist-pl and the german-pl ‘There were the fascists and the Germans’

Such a correlation has already been identified in Berruto and Cerruti (2015) for other varieties of Italian and is explicable in terms of a lesser salience of the plural semantics of lists, given the fact that each item could equally be singular or plural. To conclude, the study of sub-standard agreement in the existential construction confirms what has already been argued for the same variable based on other corpora, concerning both the social markedness of this construction and the structural factors that facilitate its realisation. However, the issue regarding the geographic markedness of this feature remains open. Such a factor should be weighed in further studies by comparing this sociolinguistic setting with other situations in which local dialects display patterns of agreement more similar to Standard Italian. 6. Conclusions The use of the ParVa corpus has enabled us to attain a broader and updated perspective on italiano popolare, considering the fact that spoken data made it possible to identify new features with the same diastratic markedness as those traditionally associated with this variety but not attested in written sources. We have also had the opportunity to test previous findings on one single variable, i.e. agreement in the existential construction, and we have demonstrated that the same factors are responsible for the emergence of sub-standard variants both in spoken italiano popolare and in other less sociolinguistically marked varieties of Italian (see Berruto and Cerruti 2015). As for the more general issue concerning linguistic simplification, the starting point of our reflection was Miestamo’s (2017) crosslinguistic account of linguistic complexity. By comparing structures occurring in Standard Italian and in italiano popolare, i.e. by applying Miestamo’s framework to intralinguistic variation as well, we have identified many of the structures occurring in the latter variety to be systematically simpler than those occurring in the standard.

82

Silvia Ballarè and Eugenio Goria

Abbreviations 1pl 1sg 3pl 3sg comp inf neg obl pl pp pres pst refl rel

1st person plural 1st person singular 3rd person plural 3rd person singular complementizer infinitive negation oblique plural past participle present past reflexive relative

References Adger, David and Jennifer Smith. 2010. Variation in agreement: A lexical feature-based approach. Lingua 120. 1109–1134.  https://doi.org/10.1016/j.lingua.2008.05.007 Alisova, Tatiana. 1965. Relative limitative e relative esplicative nell’italiano popolare. Studi di Filo­logia Italiana 23. 299–333. Ballarè, Silvia. 2015. La negazione di frase nell’italiano contemporaneo: un’analisi sociolinguistica. Rivista Italiana di Dialettologia 39. 37–71. Bentley, Delia, Francesco Maria Ciconte and Silvio Cruschina. 2013. Existential constructions in cross-linguistic perspective. Rivista di Linguistica 25 (1). 1–13. Bernini, Giuliano and Paolo Ramat. 1996. Negative sentences in the languages of Europe. BerlinNew York: de Gruyter.  https://doi.org/10.1515/9783110819748 Berruto, Gaetano. 1983. L’italiano popolare e la semplificazione linguistica. Vox Romanica 42. 38–79. Berruto, Gaetano. 1987. Lingua, dialetto, diglossia, dilalìa. In Romania et Slavia Adriatica. Fest­ schrift für Zarko Muljačić, 57–81. Hamburg: Buske. Berruto, Gaetano. 2012 [1987]. Sociolinguistica dell’italiano contemporaneo. 2nd edn. Roma: Carocci (1987: Roma: La Nuova Italia Scientifica). Berruto, Gaetano and Massimo Cerruti. 2015. Un esercizio di analisi variazionista: l’accordo verbale nel costrutto locativo-esistenziale-presentativo. In Grazia Busà and Sara Gesuato (eds.), Lingue e contesti. Studi in onore di Alberto A. Mioni, 609–620. Padova: CLEUP. Cerruti, Massimo. 2011. Regional varieties of Italian in the linguistic repertoire. International Journal of the Sociology of Language 210. 9–28. Cerruti, Massimo. 2014. From language contact to language variation: a case of contact-induced grammaticalization in Italo-Romance. Journal of Language Contact 7. 288–308. https://doi.org/10.1163/19552629-00702003

Chapter 4.  C’era i fascisti e i tedeschi 83

Cerruti, Massimo. 2016. Costruzioni relative in italiano popolare. In Federica Guerini (ed.), Ita­ liano e dialetto bresciano in racconti di partigiani, 77–116. Roma: Aracne. Cerruti, Massimo and Riccardo Regis. 2014. Standardization patterns and dialect/standard convergence: A North-Western Italian perspective. Language in Society 43 (1). 83–111. https://doi.org/10.1017/S0047404513000882 Chambers, Jack K. 2004. Dynamic typology and vernacular universals. In Bernd Kortmann (ed.), Dialectology meets typology: Dialect grammar from a cross-linguistic perspective, 128–145. Berlin-New York: de Gruyter. Cini, Monica. 2008 (ed). I verbi sintagmatici in italiano e nelle varietà dialettali. Stato dell’arte e prospettive di ricerca, Frankfurt am Mein: Peter Lang. https://doi.org/10.3726/978-3-653-01094-7 Cortelazzo, Manlio. 1972. Avviamento critico allo studio della dialettologia italiana. In Line­a­ menti di italiano popolare. Pisa: Pacini. Coseriu, Eugenio. 1980. ‘Historische Sprache’ und ‘Dialekt’. In Joachim Göschel, Pavle Ivi and Kurt Kehr (eds.), Dialekt und Dialektologie, 106–22. Wiesbaden: Steiner. Crocco, Claudia. 2015. Corpora e testi di italiano contemporaneo. In M. Iliescu and Eugeen Roegiest (eds.), Manuel des anthologies, corpus et textes romans, 509–534. Berlin: de Gruyter. https://doi.org/10.1515/9783110333138-033 De Mauro, Tullio 1970[1963]. Storia linguistica dell’ Italia unita. Seconda edizione, Roma-Bari: Laterza. Goria, Eugenio and Caterina Mauri. 2018. Il corpus KIParla: una nuova risorsa per lo studio dell’italiano parlato. In Francesca Masini & Fabio Tamburini (eds.) Club Working Papers in Linguistics. Vol. 2. Bologna: Circolo linguistico dell’Università di Bologna. Guerini, Federica (ed.) 2016. Italiano e dialetto bresciano in racconti di partigiani, Roma: Aracne. Koch, Peter and Wulf Oesterreicher. 1990. Gesprochene Sprache in der Romania: Französisch, Ita­lienisch, Spanisch. Tubinga: Niemeyer.  https://doi.org/10.1515/9783111372914 Kortmann, Bernd and Benedikt Szmrecsanyi. 2009. World Englishes between simplification and complexification. In Thomas Hoffmann and Lucia Siebers (eds.), World Englishes – Problems, properties and prospects: Selected papers from the 13th IAWE conference (Varieties of English around the world): 263–286, Amsterdam: John Benjamins Publishing. https://doi.org/10.1075/veaw.g40.17kor Manzini, M. Rita and Leonardo Savoia. 2005. I dialetti italiani e romanci. Morfosintassi generativa, Alessandra: Dell’Orso. Matras, Yaron and Jeanette Sakel. 2007, Investigating mechanism of pattern replication in language convergence. Studies in Language 31 (4). 829–865.  https://doi.org/10.1075/sl.31.4.05mat McWhorter, John H. 2001. The world’s simplest grammars are creole grammars. Linguistic Typology 5. 125–166.  https://doi.org/10.1515/lity.2001.001 Miestamo, Matti. 2008. Grammatical complexity in cross-linguistic perspective. In Miestamo Matti, Kaius Sinnemakl and Fred Karlsson (eds.), Language complexity. Typology, contact, change. Amsterdam-Philadelphia: John Benjamins Publishing. https://doi.org/10.1075/slcs.94.04mie Miestamo, Matti. 2017. Linguistic diversity and complexity. Lingue e Linguaggio vol XVI(2): 227–253. Sanga, Glauco. 1980. Lettere dei soldati e formazione dell’italiano popolare unitario. In La grande guerra. Operai e contadini lombardi nel primo conflitto mondiale, 43–65. Milano: Silviana.

84

Silvia Ballarè and Eugenio Goria

Tagliamonte, Sali A. 2009. There was universals; Then there weren’t: A comparative sociolinguistic perspective on ‘Default Singulars’. In Markku Fillpula, Juhani Klemola and Heli Paulasto (eds.), Vernacular universals and language contacts: Evidence from varieties of English and beyond, 103–129. New York-Oxford: Routledge. Tagliamonte, Sali A. and R. Harald Baayen. 2012. Models, forests and trees of York English: Was/ were variation as a case study for statistical practice. Language Variation and Change 24 (2). 135–178.  https://doi.org/10.1017/S0954394512000129

Chapter 5

Language change caught in the act A case study of Frisian relative pronouns Jelske Dijkstrai,ii, Wilbert Heeringai, Emre Yılmaziii, Henk van den Heuveliii, David van Leeuweniii and Hans Van de Veldei iFryske

Akademy, Leeuwarden, iiMercator European Research Centre on Multilingualism and Language Learning / iiiRadboud University, Nijmegen

This study investigates language change in Frisian relative pronouns in a corpus of Frisian radio broadcasts (1966–2015). In spite of the limitations of this corpus, we were able to catch language change in the act. The analyses show that until the 1980s the younger speakers in these broadcasts lead the rise of t-full relative pronouns, a change that was first observed in literature at the end of the 19th century. From the 1980s onwards the new younger generation reversed this change and increasingly started using the t-less relative pronouns. Additionally, the increase of t-less forms occurred mostly in non-scripted, spontaneous speech. It does not seem to play a role whether the speaker is a presenter or an interviewee/guest. Keywords: Audience Design, scriptedness, Frisian, relative pronouns, t-deletion, radio, language change, real time, apparent time

1. Introduction This real time study examines language change in progress in Frisian, a minority language spoken in the province of Fryslân in the north of the Netherlands.1 The study is conducted using a corpus of audio fragments from radio broadcasts aired between 1966–2015. It documents and analyses the change of four Frisian relative pronouns that end in /t/, i.e. dy’t ‘who/that’, dêr’t ‘where’, wêr’t, ‘where’ and wa’t ‘who(m)’, to their t-less counterparts, i.e. dy, dêr, wêr and wa. Our study investigates the role of factors related to the radio medium, such as the scriptedness of the speech, whether the speaker is a professional radio presenter or not, the role of 1. The FAME! Project is funded by the Netherlands Organisation for Scientific Research NWO, Project 314-99-119 (Frisian Audio Mining Enterprise). https://doi.org/10.1075/silv.22.05dij © 2019 John Benjamins Publishing Company

86 Jelske Dijkstra et al.

traditional sociolinguistic factors such as gender, age (apparent time) and period (real time). For the realisation of final /t/ in these relative pronouns in spoken Frisian two opposing tendencies are mentioned in the literature. Brouwer (1959) claimed that the use of the t-full forms of these relative pronouns was increasing at that time in spoken Frisian and thirty years later Van der Meer (1991) noticed that the use of the t-less forms was increasing. As we only have these observations without any quantitative support and no information at all on possible sociological or linguistic factors, we wonder whether we will be able to capture this reversal. Our first two hypotheses are therefore: 1. Newer broadcasts show more t-less forms than older ones. 2. Younger speakers show more t-less forms than older ones from the moment the change is reversing. According to Labov’s (2001: 293) gender paradox women conform more closely than men to sociolinguistic norms that are overtly prescribed, but conform less than men when they are not. This leads to a third hypothesis: 3. If a gender effect shows up, we might expect that in the initial stage of the change towards t-less relative pronouns, women show higher rates of t-less forms than men, as there is no clear standard for spoken Frisian (Bloemhof, De Haan and Versloot 2013: 722). Speech used by presenters and other speakers in radio broadcasts show many variations in style, depending on the addressee, topic, genre and setting of the broadcast. In the Audience Design model Bell (1984: 158, 2001: 141) described how radio presenters adapt their speech to the speech of their intended audience and/or opt for a variety that is acceptable for (most members of) their target audience. He compared the realisation of several linguistic variables by the same presenters whilst reading the news at a local and a national radio broadcasting station. Their speech on national broadcasts contained around twenty percent more standard variants compared to the local broadcasts. Since other factors such as setting, genre and speaker were constant in his studies, the shift in speech style reflected their perception of the language norms and variation in the speech community. Another strategy within style variation is Speaker Design where the style shift is a source of active creation, presentation and recreation of speaker identity (Bell 2001: 147; Coupland 1985: 168, 2001: 366; Schilling-Estes 2002: 388). This style shift is not reactive or responsive as in Audience Design, but rather proactive or initiative. Several studies have been conducted with the Audience Design and/or Speaker Design approach (e.g. Bell 1982, 1984, 1999, 2001; Bell and Johnson 1997; Coupland 1985, 2001; Cutillas-Espinosa and Hernández-Campoy 2006; Cutillas-Espinosa, Hernández-Campoy and Schilling-Estes 2010; Hernández-Campoy and Cutillas-Espinosa 2010, 2012;



Chapter 5.  Language change caught in the act 87

Hernández-Campoy and Jiménez-Cano 2003; Schilling-Estes 1998, 2002). The current study uses audio fragments from broadcasts aired by the regional broadcaster Omrop Fryslân. As the presenters of this radio station are instructed to use a register similar to colloquial speech (Tolsma 2009: 2), we expect that in our study professional broadcasters converge towards the speech of their audience and interlocutors, in line with the Audience Design model. This leads to the fourth hypothesis of our study: 4. Professional presenters will not differ in the use of t-full or t-less forms from their guests or interviewees. Cutillas-Espinosa and Hernández-Campoy (2006) showed that other factors can also play a role in style shifting. In their study, the presenter of a radio broadcasting station in Murcia used more standard (Castilian Spanish) variants in the broadcasts whilst his listeners who called in used more non-standard (Murcian Spanish) forms. Here the presenter clearly diverged his speech from his audience. During an interview with the two Murcian-speaking scholars the presenter mostly used non-standard Murcian Spanish forms in his speech. The presenter further admitted during the interview that he did not like using the standard accent at work. Therefore the authors concluded that the presenter’s style-shift did not meet either the Audience, or the Speaker Design approaches. They argued that the shift was due to a script, that was not freely chosen by the presenter and probably written by someone else complying with the language policy of the radio station (Cutillas-Espinoza and Hernández-Campoy 2006: 325). In scripted speech the influence of orthography and written norms are higher than in non-scripted speech. Present-day Frisian grammar books prescribe the t-full forms (see also the section on relative pronouns). Consequently, our fifth hypothesis is as follows: 5. The number of t-full forms will be higher in scripted than in non-scripted speech. Several studies have been conducted on language variation and change using radio broadcasts. For example, Van de Velde, Van Hout and Gerritsen (1997) used archived radio broadcasts (1935–1993) to examine phonological variation and change in Dutch spoken in the Netherlands and Flanders. They observed phonetic divergence between both varieties of standard Dutch, as spoken by radio presenters. Hernández-Campoy and Jiménez-Cano’s (2003) real time study showed increasing diffusion of standard Castilian features in non-standard Murcian Spanish during interviews broadcast from 1975–2000. In both studies the authors argue that radio broadcasts are an excellent source for studying language variation and change. However, they also point out the limitations of this medium, i.e. the lack of information on social characteristics of the speakers, an unequal distribution of variables over different topics, genres and/or settings, the speakers’ representativeness of

88

Jelske Dijkstra et al.

their speech community, social pressure that urges the speaker to be linguistically correct, etc. (Van de Velde et al. 1997: 367; Hernández-Campoy and Jiménez-Cano 2003: 342). 2. Omrop Fryslân Omrop Fryslân (Fryslân Broadcasting Corporation) is the regional radio (and television) broadcaster in Fryslân and Frisian is prescribed as their official language (Tolsma 2009: 2). Consequently, their target audience is defined as those who understand Frisian. Promotion of the Frisian language is not the main goal of Omrop Fryslân, and it is not motivated by linguistic purism. The style book of Omrop Fryslân states: “It Frysk sa’t wy dat brûke by radio en televyzje moat oanslute by de omgongstaal fan de minsken, it moat natuerlik oerkomme en gjin ôfstân skeppe” ‘The Frisian that we use on radio and television must be similar to people’s colloquial speech, it should be natural and not create distance’ (Tolsma 2009: 2). This means that archaic words and phrases are avoided as much as possible. However, Dutch loanwords are often replaced with a Frisian equivalent if this word is not too artificial (Tolsma 2009: 3). Omrop Fryslân uses Frisian as much as possible. Only if the guest/ interviewee does not understand Frisian, the presenter of Omrop Fryslân switches to Dutch. The guests/interviewees can use the language they choose whether it be Dutch or one of the Frisian or mixed Frisian-Dutch dialectal varieties spoken in Fryslân (Tolsma 2009: 3). In practice, this policy implies that mainly Frisian is used in the radio programs, but also Dutch or a dialectal variety are allowed. Being able to speak Frisian is a requirement for presenters, editors and producers. Presenters and editors are expected to have a basic level of proficiency (in both speaking and writing) in Frisian (Tolsma 2009: 4). Omrop Fryslân has a rich history, especially in radio broadcasting. The precursor of Omrop Fryslân started broadcasting just after World War II, and was originally the regional station of a broader area in the North of the Netherlands, using mainly Dutch (the language shared by the whole area) and to a smaller extent Frisian. At that time, most of the broadcasts were pre-recorded and scripted. The first live coverage in Frisian was in 1972. In 1978 the broadcasting station was split into three regional broadcasters, i.e. Radio Noord (Radio North) for the provinces Groningen and Drenthe, Radio Oost (Radio East) for the provinces Overijssel and Gelderland, and Radio Fryslân (Radio Fryslân) for the province of Fryslân, the latter using predominantly the Frisian language. Since 1988 the Frisian broadcasting station operates as an independent radio station. In 1994 they also started daily television and the station’s name changed to Omrop Fryslân (“Skiednis Omrop Fryslân” n.d.).



Chapter 5.  Language change caught in the act 89

3. Frisian in the Netherlands Frisian is an endogenous minority language spoken in the province of Fryslân, the Netherlands. In this region Frisian was recently appointed as the second official language, Dutch being the first one (“Wet Gebruik Friese Taal” 2014). About 75% of the inhabitants of Fryslân (population 647,071 in 2017, Statline) speak Frisian, and 55% of the inhabitants are native speakers of Frisian (Provinsje Fryslân 2015: 10). There is no standard for spoken Frisian (Bloemhof, De Haan and Versloot 2013: 722). There is however a codified written standard with uniform spelling rules. Written Frisian contains numerous archaic lexical elements that are infrequently used in spoken Frisian. Since Dutch is not only the dominant language in the Netherlands at large, but also the main language used in education, all speakers of Frisian are bilingual. Frisians are predominantly literate in Dutch: only 15% of the Frisians indicate they can write Frisian well (Provinsje Fryslân 2015: 10). Frisian and Dutch are both West-Germanic languages and share many grammatical and lexical features. Furthermore, due to language contact, the Frisian language has a growing number of Dutch loanwords and interferences on the phonological, morphological and syntactic levels (De Haan 1997). 4. Relative pronouns The current paper studies the replacement of t-full forms of four Frisian relative pronouns, i.e. dy’t ‘who/that’, dêr’t ‘where’, wêr’t ‘where’ and wa’t ‘who(m)’, by their t-less counterparts. Table 1 gives an overview of all relative pronouns in present-day Frisian and Dutch. The Frisian relative pronoun dy’t and its Dutch counterpart die are used when the antecedent is a singular masculine, a singular feminine or a plural, see (1) and (2) in Table 1. When the antecedent is neuter, the relative pronoun dat ‘that’ is used in both languages, see Example (3) (“Taalportaal | Relative pronouns” 2018). The Frisian wa’t ‘who(m)’ and Dutch wie (4) are used as free relatives referring to a person, and not to an object or animal. In case the antecedent is a location, the relative pronouns are dêr’t ‘where’ or wêr’t ‘where’ in Frisian, and waar in Dutch (“Taalportaal | Relative pronouns” 2018). The Frisian wêr’t is used in free relatives only, in other cases dêr’t is used, see Examples (5) and (6). Due to the influence of Dutch, dêr’t is often substituted by wêr’t (De Haan 2001: 183–184; “Taalportaal | Relative pronouns” 2018). Further, wat ‘which’ can be used in the case of free relatives in both Frisian and Dutch, see also (7) (“Taalportaal | Relative pronouns” 2018).

90 Jelske Dijkstra et al.

Table 1.  Overview of relative pronouns in Frisian and Dutch Frisian 1 2 3 4 5 6 7

dy’t

Dutch

de man/frou dy’t it gers meant dy’t [dit] de doarren dy’t read binne dat [dɔt] it hûntsje dat no blaft wa’t [ʋa:t] wa’t dat dien hat, is in held dêr’t [dɛ:t] de grûn dêr’t er op stiet wêr’t [wɛ:t] ik wit net wêr’t it boek is wat [ʋɔt] hy is fuort, wat ik net ferwachte hie [dit]

English translation

die

[di]

die

[di]

dat

[dat]

wie

[ʋi]

waar [ʋa:r] waar [ʋa:r] wat

[ʋat]

de man/vrouw die het gras maait de deuren die rood zijn het hondje dat nu blaft wie dat heeft gedaan is een held de grond waar hij op staat ik weet niet waar het boek is hij is weg, wat ik niet had verwacht

‘the man/woman who is mowing the grass’ ‘the doors that are red’ ‘the little dog that is barking now’ ‘whomever has done that, is a hero’ ‘the land he is standing on’ ‘I do not know where the book is’ ‘he is gone, which was not what I expected’

As shown in Table 1, there are four Frisian relative pronouns, i.e. dy’t ‘who/that’, dêr’t ‘where’, wêr’t ‘where’, and wa’t ‘who’, that end in a clitic /t/, which is spelled as ’t. Their Dutch counterparts lack this clitic. Like Dutch, Frisian has degemination: this means that dat + t or wat + t is always realised as [dɔt] or [ʋɔt] (see (3) and (7) in Table 1), and in speech cannot be distinguished from the forms without clitic /t/. An ending in ’t is not uncommon in Frisian. Other conjunctions, such as hoe’t ‘how’, doe’t ‘when’, wannear’t ‘when’, also exhibit a clitic /t/ at the end (“Taalportaal | Relative pronouns” 2018). The /t/-ending in the four above mentioned relative pronouns is generally seen as a reduced form of dat ‘that’ or oft ‘if ’ (Hoekstra 2002: 64; Van Coetsem 1960: 327; Van der Meer 1991: 45). However, Van der Woude (1960: 337) questions whether ’t is derived from dat, since he mainly found conjunctions with or without ’t and barely any conjunctions with the full form dat in Middle Frisian texts. The Frisian grammar of Van Blom (1889: 772) is the first one that mentions the reduction of dat as a conjunction to ’t; not in the section on relative pronouns (where only forms and examples without ’t are given), but in the section on conjunctions with a remark that it is a linking sound before vowels. The same holds for Postma and De Clercq’s grammar (1904: 96). In Sytstra and Hof ’s Frisian grammar (1925: 116–117), the forms dy’t and hwa’t (former spelling of wa’t) are used as optional within the section on relative pronouns for the first time, but without any further explanation. They also mention dêr’t and hwer’t (former spelling of wêr’t) (Sytstra and Hof 1925: 52). Fokkema (1948: 55, 1967: 51) prescribes the forms dêr’t and hwer’t with ’t as relative pronouns. For dy and hwa Fokkema (1948: 54, 1967: 50–51) states that they are often followed by ’t. On the other hand, Sipma (1949: 78–79) mentions the forms dy’t and hwa’t only, not their

Chapter 5.  Language change caught in the act 91



t-less counterparts. He further states that the form hwa dat is also still in use. Additionally, his grammar prescribes dêr’t and wêr’t as the only forms used (Sipma 1949: 79). From that time onwards, the t-full forms appear consistently in Frisian grammars (Tiersma 1999; Popkema 2006; “Taalportaal | Relative pronouns” 2018). In the middle of the 20th century Brouwer (1959: 64) still observes an increase in the use of the t-full forms of these relative pronouns and other conjunctions in spoken Frisian. However, by the end of the 20th century, Van der Meer (1991: 44) notices that the use of t-full forms of relative pronouns and other conjunctions was decreasing in Frisian speech. He also observes that this trend is not restricted to the younger generations. The Syntactic Atlas of the Dutch Dialects (Barbiers et al. 2006) shows that both the t-full and the t-less forms of the relative pronouns in this study are used in Frisian. Due to the lack of quantitative analyses and reliable information on the socio-geographic distribution of the phenomenon, it remains unclear whether this is a stable (stratified) variation pattern or an abortion and reversal of a change in progress. In this paper we will investigate the realisation of clitic /t/ in these relative pronouns in real time over the past fifty years (1966–2015) in an attempt to get more insight into this phenomenon. 5. Method 5.1

The data

The data used for this study come from the FAME! Speech Corpus (Yılmaz et al. 2016: 1537). This corpus comprises of 203 radio fragments that are extracted from various programs broadcast by Omrop Fryslân between 1966 and 2015. The corpus was collected as a training and testing set for the development of a speech recogniser that can handle Dutch and Frisian simultaneously. The duration of each fragment is approximately five minutes, the total corpus consists of 18.5 hours of speech. All audio fragments were transcribed and annotated by two native speakers of Frisian and checked and corrected by the first author, who is also a native speaker of Frisian. For the purpose of this study, only utterances of Frisian were used. Further, the utterances of non-native speakers of Frisian were discarded by the first author, based on accent or pronunciation, intonation, word use, and fluency.2 All fragments were annotated as scripted (read aloud), semi-scripted (some pre-annotated 2. See also Amino and Osanai (2014); Van Maastricht, Krahmer and Swerts (2016), and Vieru, Boula de Mareüil and Adda-Decker (2011) for work on judgments of non-native speech by native speakers.

92

Jelske Dijkstra et al.

phrases or words), and non-scripted (spontaneous). These three styles were distinguished on the basis of program genre and features such as prosodic cues, rhythmic irregularities, pauses, fillers, and false starts of the speaker (Cucchiarini, Strik and Boves 2002; Dellwo, Leemann and Kolly 2015; Levin, Schaffer and Snow 1982). Contexts where the relative pronoun was followed by a /t/, such as dy’t thús binne ‘that are at home’, and contexts with an inflectional conjunction such as dêr’tst wennest ‘where-you live’, were excluded from the analysis, as it is impossible to decide whether the clitic /t/ is pronounced. Table 2.  Number of tokens and percentages of (t), split up for (dy’t), (dêr’t), (wêr’t), (wa’t) and the Dutch interference (wie’t) in scripted, semi-scripted and non-scripted radio fragments (n = 776)  

Scripted

Semi-scripted

Non-scripted

Total

 

 

n

%

n

%

n

%

n

%

(dy’t)   (dêr’t)   (wêr’t)   (wa’t)   *(wie’t)   Total

dy’t dy dêr’t dêr wêr’t wêr wa’t wa wie’t wie  

191  10  44   1  15  10   7   0   0   0 278

 95.0   5.0  97.8   2.2  60.0  40.0 100.0   0.0   0.0   0.0  

57  3  6  0  9  7  4  0  0  0 86

 95.0   5.0 100.0   0.0  56.2  43.8 100.0   0.0   0.0   0.0  

202  89  19   2  54  41   1   0   1   3 412

 69.4  30.6  90.5   9.5  56.8  43.2 100.0   0.0  25.0  75.0  

450 102  69   3  78  58  12   0   1   3 776

 81.5  18.5  95.8   4.2  57.3  42.7 100.0   0.0  25.0  75.0  

Table 2 presents the number of tokens of the relative pronouns under investigation in the final dataset. The data show that the Frisian relative pronoun wa’t was occasionally replaced by its Dutch counterpart wie, even once produced with the (Frisian) clitic /t/: wie’t. Therefore we included these tokens in the variable (t). Forms with dat, e.g. dy dat or dêr dat, were not found in the data. It should be noted that (t) has a rather low frequency: 776 tokens in 18.5 hours of recordings, i.e. an average of less than one per minute. Furthermore, the number of tokens is spread over a large number of speakers (total of 266 speakers, of which 162 appear only once, since some appear multiple times in the corpus) and the number of observations per speaker ranges between 1 and 12. For 295 speakers there were no tokens of (t) at all. This low frequency, and the unequal spread of the tokens over the speakers in this corpus, hampers an analysis of a large set of internal and external factors.

Chapter 5.  Language change caught in the act 93



5.2

Procedure

All tokens of (t) in the target words (n = 776), i.e. dy’t, dêr’t, wêr’t, wa’t, and the Dutch interference wie’t, were coded (absence or presence of [t]), except those followed by a /t/, e.g. de doar dy’t tydlik op slot sit ‘the door that is temporarily locked’. Speaking rate was automatically measured, with a manual check of the number of syllables afterwards by the first author, since the script used for counting the syllables was based on Dutch syllabification rules, resulting in several corrections. Also the right hand environment of (t) was coded (factor linguistic context). Age of the speakers was determined by means of searches for their birth year on the internet, in the provincial library and in family archives. For 76% of the speakers the birth year could be traced, and the age of the speaker in the year of the broadcast was calculated. For the remaining speakers age was estimated on the basis of vocal features and the content of the audio fragment. 5.3

Analysis

The data were analysed using the Cumulative Link Mixed Model (CLMM) in R (The R Foundation for Statistical Computing, http://CRAN.R-project.org) by applying the CLMM function in the Ordinal package (Christensen 2015). The ordered dependent variable was (t) with 0 for the t-less relative pronoun and 1 for the t-full relative pronoun. Three factors were entered as random intercepts. Each speaker was entered as such, so the data were controlled for individual variance since some speakers appeared multiple times in the data. The variable Pronoun i.e. (dy’t), (dêr’t), (wêr’t) or (wa’t) was also entered as random intercept to control for this linguistic factor. Entering this factor as a fixed factor resulted in an invalid model. This can be explained by the unequal distribution: on average, dy’t was used 3.9 times per speaker, the other pronouns between 1 and 1.8. Finally, speech rate was inserted as random intercept so the data were also controlled for this variable. Fixed factors were year of broadcast, age of speakers, type of speaker (presenter of Omrop Fryslân, interviewee/guest), gender, degree of scriptedness (scripted, semi-scripted, non-scripted) and linguistic context. Pronouns were followed by 29 different linguistic contexts (i.e. right hand environment of the pronoun), but we recoded them to three values: vowel, consonant and pause (including ‘eh …’). We also included the interaction between year of broadcast and the age of speakers. These factors were entered into the initial model. The Akaike Information Criterion (AIC) was 593.34. With backward analysis we obtained the best possible model with an AIC of 587.3 which is a significant improvement compared to the AIC of the initial model. The best model is presented in the Results section. Note that both models were applied to the data of all speakers (presenters and guests/interviewees) and all types of scriptedness of the speech.

94 Jelske Dijkstra et al.

In the statistical analyses the age of the speakers and the year of broadcast were used, however, for the sake of clarity the mean scores of (t) for three age groups (20–41, 42–63 and 64–82) and three periods (1966–1982, 1983–1999, 2000–2015) will be presented in the figures below. It should be noted that the dataset has some limitations that have consequences for the statistical analysis and interpretation of the results. We already mentioned the low frequency of (t) (n = 776), the large number of speakers (N = 266) and the fact that the token frequency per speaker is very low, hampering an analysis of the language internal factors. Furthermore, there is almost no scripted and semi-scripted speech for the oldest age group. This type of speech is predominantly produced by professional broadcasters, who used to retire between 60 and 65 years old. Consequently, interaction effects between age of the speaker and scriptedness cannot be tested. However, to get a better insight, the results of (t) will be split up for the different types of scriptedness. 6. Results The best model is presented in Table 3. Year of broadcast is a significant factor in this analysis: the more recent the audio fragment, the more t-less forms are used by the speakers. However, it should be noted that the effect size is small. Table 3.  The best model for t-realisation in dy’t, dêr’t, wêr’t, wa’t and the Dutch interference wie’t (n = 776, N = 266)   Year of broadcast Age Scripted vs. semi-scripted Scripted vs. non-scripted Year:Age

Estimate

Standard error

z-value

Pr(>|z|)

−0.77  0.42 −0.39 −2.56  0.53

0.25 0.25 0.75 0.65 0.25

−3.11  1.67 −0.52 −3.95  2.11

p  [ˈkanne]), and permutation of liquid consonants (mujer ‘woman’: [muˈχer]>[muˈχel]; algo ‘something’: [ˈalgo]> [ˈargo]). The correlation of linguistic variables with social variables (groups) and time cohorts, allowed Hernández-Campoy and Jiménez-Cano (2003) to compare and analyse the sociolinguistic behaviour of the different social groups in terms of standardisation (adoption of Castilian Spanish features) or non-standardisation (maintenance of local Murcian features) and their tendencies diachronically. The results (Figure 2) showed a slow but steady monotonic pattern of approximation to the standard Castilian Spanish prestige model, with the subsequent attrition of local vernacular features. This means that the pronunciation of Murcian speakers was closer to the standard in the year 2000 (81%) than in 1975 (57%). 90

Group 1 (Politicians)

85

Group 2 (Non-politicians)

80 75 (%)

70 65 60 55 50 45 40

1975–1979

1980–1984

1985–1989

1990–1994

1995–2000

Figure 2.  Expansion of standard Castilian Spanish in Murcia per variables and groups of informants (G1: male politicians; and G2: male non-politicians). Source: adapted from Hernández-Campoy and Jiménez-Cano (2003: 335, Figure 5)

Also, although the two social groups exhibited different percentages of standard features, their evaluation of the two variants (standard and non-standard) was the same: both changed their pronunciation in exactly the same direction, increasing the percentage of high-status, standard Castilian Spanish forms in their speech.

Chapter 6.  Virtual sociolinguistics 107



As Figures 3–4 show, the general tendency was for the individual variables to converge towards standard Castilian Spanish, though to different degrees, since the adoption of standard forms was diachronically more accelerated in the prestige higher social group (Group 1) than in the lower one (Group 2): variables whose standard variant appeared to be categorical – final /r/, final /l/, para and liquids permutation – or almost categorical (/d/) in the speech of Group 1 at the beginning of the 1980s were not adopted to the same degree in the pronunciation of Group 2 until the end of the same decade. 100

Final/s/ /d/ Liquidspermutation Final/r/ para Final/l/ Assimilation

90 80 70 (%)

60 50 40 30 20 10 0

1975–1979

1980–1984

1985–1989

1990–1994

1995–2000

Figure 3.  Expansion of standard Castilian Spanish in Murcia among Group 1 (politicians). Source: Hernández-Campoy and Jiménez-Cano (2003: 335, Figure 3) 100

Final/s/ /d/ Liquidspermutation Final/r/ para Final/l/ Assimilation

90 80 70 (%)

60 50 40 30 20 10 0

1980–1984

1985–1989

1990–1994

1995–2000

Figure 4.  Expansion of standard Castilian Spanish in Murcia among Group 2 (male non-politicians). Source: Hernández-Campoy and Jiménez-Cano (2003: 334, Figure 4)

108 Juan Manuel Hernández-Campoy

There is, therefore, a group of linguistic variables which are prone to standardisation – variables (r), (l), (para) and consonant permutation. The linguistic changes shifting from non-standard forms usage to standard ones captured here in their terminal process were the cases of intervocalic /r/ in the word para and consonant permutation (r>l and l>r). The standard forms of the postvocalic /l/ and /r/ variables in word-final position were not well embedded until the mid-1990s. Intervocalic /d/ is at an advanced stage of standardisation and behaves like a sociolinguistic marker (subject to both social and stylistic variation). There is another group which is considerably more reluctant to standardise: final /s/ and consonant assimilation. In fact, at least in Peninsular Spanish, according to Martínez-Martín (1983), the process of regressive consonant assimilation of consonantal clusters – a salient feature of non-standard Spanish varieties in the south – is described as a phenomenon in expansion in apparently standard Castilian Spanish-speaking areas in the northern regions of Spain (see also Hernández-Campoy and Villena-Ponsoda 2009; and Fernández de Molina and Hernández-Campoy 2018). As for word-final postvocalic /s/, the loss of /s/ is categorical in Murcian Spanish (see Hernández-Campoy and Trudgill 2002). Both variables are essentially southern features so deeply rooted within the Murcian speech community that they have become part of the local identity. But, in addition to these sociolinguistic constraints, in the case of postvocalic /s/, this feature also affects the markers for number in nouns (casa-casas: ‘house/ houses’) and person in verbs (él tiene-tú tienes: ‘he has’/’you have’), which make Murcian Spanish diverge from Standard Castillian through the use of vowel quality rather than morpho-syntactic marking. The degree of dissimilarity between the linguistic systems of both dialectal areas (southern peninsular Spanish vs. northern peninsular Spanish) for these morpho-syntactic markers condition the process and pattern of diffusion of Standard Spanish – even more than language loyalty: the adoption of the Standard Castilian pronunciation for these features would be a linguistic cost that the local dialects would not be able to afford: it would mean a dramatic change in, for example, the Murcian vowel system, as /æ/, /ɛ/ and /ɔ/ would simply disappear tending to a five vowel system; and, consequently, there would also be morpho-syntactic changes (ending -s rather than vowel quality differences) to mark number in nouns and person in tenses, for example (see Hernández-Campoy 2008a and Hernández-Campoy and Villena-Ponsoda 2009).

Chapter 6.  Virtual sociolinguistics 109



3. Panel study with historical corpora of written correspondence The development of electronic linguistic corpora as linguistic sources (see Bauer 2002; Schneider 2002; and Cantos 2012) is allowing Historical Sociolinguistics to study remote periods of a language and its users’ sociolinguistic behaviour (see Nevalainen and Raumolin-Brunberg 1996, 2003; Hernández-Campoy and Conde-Silvestre 2012). In this sense, the relevance of official and private historical written correspondence has meant, for example, a crucial contribution to trace language variation and change in the past as well as their social origin and direction (see Nevala and Palander-Collin 2005; Nevalainen and Tanskanen 2007; Palander-Collin 2010; Conde-Silvestre and Hernández-Campoy 2013). Similarly, letters have also allowed the reconstruction of ancient community values as reflected in the written communicative interaction through language choice and use and the subsequent transmision of linguistic as well as social meaning (see Hernández-Campoy and García-Vidal 2018b). Conde-Silvestre and Hernández-Campoy (2013) made use of a corpus of private written correspondence as linguistic material to investigate the diffusion of a change in progress in medieval times. As for the archival source, the Paston Letters is a collection of 422 authored documents (246,353 words) written by 15 male and female members belonging to different generations of this minor gentry Norfolk family from 1425 to 1503 (Table 2). The socio-historical and linguistic importance of these documents is extraordinary, as they provide us with information on the political and domestic history of 15th century England: historically, a period of great turbulence and anarchy within the framework of the War of the Roses (1455–1487), and, sociolinguistically, crucial for the development of the English language – with the implementation and diffusion of the incipient standard norm (Constable 1976; Schäfer 1996; Hernández-Campoy and Conde-Silvestre 1999; Conde-Silvestre and Hernández-Campoy 2004; or Hernández-Campoy 2008b). As far as the linguistic variable is concerned, it was an innovation in the spelling Table 2.  Paston Family members (adapted from Bergs 2005: 61; Figure 11) Generation 1 William I Agnes          

Generation 2

1378–1444 John I ?1400–1479  Margaret   Edmond I   Elizabeth   William II   Clement II    

Generation 3 1421–1466 ?1420–1484  1425–1449 ?1429–1488  1436–1496   1442–?1479 

 

John II John III Margery Edmond II Walter William III William IV

1442–1479 1444–1504 ?1455–1495   ?1443–?1504  ?1456–1479    ?1459–after1504 1479–1554

110 Juan Manuel Hernández-Campoy

practices of the period: the progressive adoption of the Roman-based new orthographic variant at the expense of the old runic symbol . The study scrutinised the use of this orthographic variants for (TH) – a sociolinguistic marker at that time (Stenroos 2004, 2006; Bergs 2007) – in the different members of the Paston Family. Conde-Silvestre and Hernández-Campoy (2013) found that the different members of the family easily illustrate the gradual but steady adoption of the new form and the subsequent extinction of the old runic throughout the 15th century, as a typical Labovian change from above (Labov 1994: 78; Conde-Silvestre and Hernández-Campoy 2013; Hernández-Campoy, Conde-Silvestre and García-Vidal 2019). Longitudinally, as reproduced in Figure 5, the incoming spreads with a positive monotonic pattern, as an example of a graphic change that projects the linguistic changes proceeding gradually, cumulatively and consecutively from generation to generation. Whereas the oldest informants show lower scores of the incoming , conversely, the youngest ones exhibit higher frequencies of the same form, so that the accumulated slope of the lines in both individual and generational terms reflects the development of the change in progress. 100 90 80 70 60 (%) 50 40 30 20 10 0

W1 1425

TH in the Pastons Totals Lineal…

J1

Ed1

W2

C2

J2

Ed2

J3

Wa

W3

W4 1503

Figure 5.  Adoption of the new form among the Paston family members (Conde-Silvestre and Hernández-Campoy 2013: 293). Members of the Paston Family are listed in the x axis using abbreviations/number for names/generation

Individually, the contrast between John I and his son John III offers an illustrating example of the gradual progression of this change within the Paston Family. There is a clearly distinctive pattern of sociolinguistic behaviour developing over their lifetimes as for variable (TH). In this process, the period of 42 years with letters available in John III’s 60 years of life  – written since he was only 17 – provides us with rich linguistic data in the context of a privileged wide lifetime perspective. John III’s sociolinguistic behaviour thus characterises a proper lifespan change,

Chapter 6.  Virtual sociolinguistics 111



clearly showing the increasing adoption of the innovative over his lifetime in the direction of a change that was becoming well embedded within the community, especially between 1480 and 1503. However, his father’s letters (John I) were written some decades before (between 1444 and 1465) and afford a more limited lifetime perception (21-year cohort). But they anyway show that, in his generation, the change was at much earlier stages of advancement and thus not so generally extended throughout the speech community yet. Although, admittedly, John I preserves a steady and high frequency of the innovative variant throughout his lifetime (Figure 6), he does not exhibit the positive progression of this change so overtly. 100

John Paston I John Paston III

95 90 (%)

85 80 75 70

1440 1445 1450 1455 1460 1465 1470 1475 1480 1485 1490 1495 1500

Figure 6.  Progress in the adoption of the innovative spelling in John I and John III: Global percentages

Age is also a determinant factor here between father and son even for their range of variation. As shown in Hernández-Campoy and García-Vidal (2018b) through a microscopic analysis of intraspeaker variation, although both John I and John III exhibit an extensive accommodative competence when addressing their correspondents according to rank, John I is subject to more variability and has a wider range of stylistic variation than his son. With an average of 80% in standardness, John I’s highest use of the innovating form is 100% with Royalty and his lowest is 25% in a draft (Letter #44, probably addressed to the Sheriff of Norfolk in 1452), showing a range of variability of 75% (Figure 7). By contrast, with an average of 89% in standardness, this linguistic change was more socially advanced in his son’s generation: John III’s highest use of the innovating form is also 100% with Royalty but, nevertheless, his lowest score is 50% in Letter #324 (addressed to his mother, Margaret Paston, in 1465), exhibiting a narrower range of variability (50%). These differences suggest a shift from heterogeneously standard (John I) to homogeneously standard (John III) in the process of expansion of the language change in progress.

112 Juan Manuel Hernández-Campoy

100 90 80 70 60 (%) 50 40 30 20 10 0

Range of variation in JP1 and JP3 Average

JP1 Highest score

JP3 Lowest score

Figure 7.  Internal variation in John I and John III in percentages

Their level of innovativeness as for variable (TH) in their sociolinguistic behaviour was basically due to generational differences and the more advance stages of the linguistic change taking place – which was affecting more overtly and intensively to John III, as well as to other Paston members of the third generation (Figure 5). This generational change was thus becoming part of the community norm and behaving as a historical change embedding within the community through its gradual, accumulative and successive process, from generation to generation, as detected through a corpus-based panel longitudinal approach (see Hernández-Campoy fc). 4. Conclusion The methodological conclusions reached from these two studies suggest that archival resources used offer a privileged perspective for the historical reconstruction of language change processes in present-day or remote languages, as virtual approaches where the problem of real time in the longitudinal research procedure is conveniently neutralised. Archived radio recordings and historical corpora of written correspondence are excellent source materials for both cross-sectional and longitudinal approaches to language variation and change. Additionally, this methodological advance can also help solve the problems of interpretation of general age-related patterns in age-grading and apparent time processes and the risk of misinterpretation of demographic change with actual linguistic change in population, as claimed by Labov (1994) or G. Sankoff (2006a, b). Historical corpora of private correspondence  – with letters written by members of the same family over several generations followed through panel surveying – ensure that the sociolinguistic diffusion of changes in the course of time can be traced, and attempts can also be made to track individual and community behaviours in this respect.



Chapter 6.  Virtual sociolinguistics 113

Obviously, as stated in Hernández-Campoy and Jiménez-Cano (2003: 342), radio recordings are subject to some limitations, such as, for example, the lack of information on the social characteristics of the informants sometimes, their representativeness for the speech community, the conditioning social pressures to being linguistically appropriate and exemplary in public situations, or the impossibility to carry out panel surveying. But it is also true that, even though the language of politicians is not fully representative of the whole speech community for different reasons (audience design, linguistic market, professional role, context, etc.), and given that no other informants’ language has usually being so much available and traceable back to the past so far, its diachronic and synchronic study may help us to reach interesting insights on the sociolinguistic behaviour of the speech community and its evolution. In fact, according to Van de Velde, Van Hout and Gerritsen (1997: 362–363), the potential use of radio broadcasting archives as linguistic data corpora requires some conditions to be fulfilled: (i) technically, the quality of the preserved recordings and the availability of their storing magnetic hardware; (ii) practically, radio language has to be appropriate for the study of variation and change in standard and/or non-standard varieties; (iii) methodologically, the combination of historical linguistic (diachronic) and sociolinguistic (synchronic) approaches for studying change in progress; and (iv) procedurally, results must validate quantitative Labovian techniques as appropriate for the study of variation and change in standard and/or non-standard varieties. In the case of historical corpora of written correspondence, as stated in Hernández-Campoy and García Vidal (2018a: 23), the reconstruction of sociolinguistic information comes from the internal evidence afforded by the preserved letters themselves, which allows us to speculate on possible reasons why the members of a family had their written practices in the way they had; and all this is based on the fact that, admittedly, as Bergs (2005: 71) underlines, this is an exercise of socio-historical reconstruction where the non-existence of evidence does not allow for conclusions about the non-existence of individual facts. This is, in fact, the cross that Historical Sociolinguistics bears: when compared to the rigorous empirical tradition of variationist research, Historical Sociolinguistics has been blamed for the lack of representativeness and its questioned validity, which made Labov (1994: 11) assert that this discipline constitutes “the art of making the best use of bad data” (see also Hernández-Campoy and Schilling 2012). These difficulties in the historical sociolinguistic practice have largely been overcome thanks to the assistance of corpus linguistics and social history, which have undoubtedly provided the discipline with empirical strength and historical basis respectively (see Bauer 2002; Schneider 2002; or Cantos 2012). In any case, it is well-known that developments in computing technology over the last twenty years have radically transformed linguistic research and that the

114 Juan Manuel Hernández-Campoy

compilation of large electronic corpora (both computer-driven and research-driven) can be crucially instrumental in the application of longitudinal studies virtually. These virtual approaches with the use of archival sources are facilitating the study of language variation and change and allowing to progress in understanding their mechanisms and motivations more accurately, more rapidly, and thus less timeand life-consuming.

References Bauer, Laurie. 2002. Inferring variation and change from public corpora. In J. K. Chambers, P. Trudgill and N. Schilling-Estes (eds.), The handbook of language variation and change, 97–114. Oxford: Blackwell. Bayley, R. 2002. The quantitative paradigm. In J. K. Chambers, P. Trudgill and N. Schilling-Estes (eds.), The handbook of language variation and change, 117–141. Oxford: Blackwell. Bergs, Alexander. 2005. Social networks and historical sociolinguistics. Studies in morphosyntactic variation in the Paston Letters (1421–1503). Berlin-New York: de Gruyter.  https://doi.org/10.1515/9783110923223 Bergs, Alexander. 2007. Spoilt for choice? problem < þe> in Peterborough Chronicle. In Alexander Bergs and Janne Skaffari (eds.), The language of the Peterborough Chronicle, 45–56. Bern: Peter Lang. Bowie, David and Malcah Yaeger-Dror. 2016. Language change in real time. In P. Honeybone and J. Salmons (eds.), Handbook of historical phonology, 603–618. Oxford: Oxford University Press. Cantos, Pascual. 2012. The use of linguistic corpora for the study of linguistic variation and change: Types and computational applications. In J. M. Hernández-Campoy and J. C. CondeSilvestre (eds.), The handbook of historical sociolinguistics, 99–122. Malden: Wiley-Blackwell. https://doi.org/10.1002/9781118257227.ch6 Conde-Silvestre, J. C. and J. M. Hernández-Campoy. 2004. A sociolinguistic approach to the difussion of chancery written practices in late fifteenth century private correspondence. Neuphilologische Mitteilungen 105 (2). 133–152. Conde-Silvestre, J. C. and J. M. Hernández-Campoy. 2013. Tracing the generational progress of language change in fifteenth century English: The diffusion of in the Paston Letters. Neuphilologische Mitteilungen 114 (3). 279–299. Constable, Giles. 1976. Letters and letter-collections. Turnhout: Brepols. Cukor-Avila, P. and G. Bailey. 2018. The effect of small Ns and gaps in contact on panel survey data. In S. E. Wagner and I. Buchstaller (eds.), Panel studies of variation and change, 210–245. New York: Routledge. Eckert, Penelope. 1997. Age as a sociolinguistic variable. In F. Coulmas (ed.), The handbook of socio­linguistics, 151–167. Oxford: Blackwell. Fernández de Molina, Elena and J. M. Hernández-Campoy. 2018. Geographic varieties of Spanish. In Kimberly Geeslin (ed.), The Cambridge handbook of hispanic linguistics, 496– 528. Cambridge: Cambridge University Press. Gertenberg, A. and A. Voeste (eds.) 2015. Language development: The lifespan perspective. Amsterdam-Philadelphia: John Benjamins Publishing.  https://doi.org/10.1075/impact.37



Chapter 6.  Virtual sociolinguistics 115

Gregersen, Frans, Marie Maegaard and Nicholai Pharao. 2009. The long and short of (æ)-variation in Danish – a panel study of short (æ) variants in Danish in real time. Acta Linguistica Haf­niensia 41. 64–82.  https://doi.org/10.1080/03740460903364086 Gregersen, Frans, Torben Jensen and Nicholai Pharao. 2018. Comparing speech samples: On the challenge of comparability in panel studies of language change in real time. In S. E. Wagner and I. Buchstaller (eds.), Panel studies of variation and change, 155–180. New York: Routledge. Hernández-Campoy, J. M. and J. C. Conde-Silvestre. 1999. The social diffusion of linguistic innovations in 15th century England: Chancery spellings in private correspondence. Cuadernos de Filología Inglesa 8. 251–274. Hernández-Campoy, J. M. and P. Trudgill. 2002. Functional compensation and southern peninsular Spanish /s/ Loss. Folia Linguistica Historica XXIII. 31–57. Hernández-Campoy, J. M. and J. M. Jiménez-Cano. 2003. Broadcasting standardisation: An analysis of the linguistic normalisation process in Murcia. Journal of Sociolinguistics 7 (3). 321–347.  https://doi.org/10.1111/1467-9481.00227 Hernández-Campoy, J.M. 2008a. Sociolinguistic aspects of Murcian Spanish. International Journal of the Sociology of Language 193. 121–138. Hernández-Campoy, J.M. 2008b. Overt and covert prestige in Late Middle English: A case study in East Anglia. Folia Linguistica Historica 29. 1–26.  https://doi.org/10.1515/FLIH.2008.1 Hernández-Campoy, J. M. and J. A. Villena-Ponsoda (2009). Standardness and non-standardness in Spain: Dialect attrition and revitalisation of regional dialects of Spanish. International Journal of the Sociology of Language 196/197. 181–214. Hernández-Campoy, J. M. and J. C. Conde-Silvestre. 2012. The handbook of historical sociolinguistics. Malden: Wiley-Blackwell.  https://doi.org/10.1002/9781118257227 Hernández-Campoy, J. M. and N. Schilling. 2012. The application of the quantitative paradigm to historical sociolinguistics: Problems with the Generalizability Principle. In J. M. Hernández-Campoy and J. C. Conde-Silvestre (eds.), The handbook of historical sociolinguistics, 65–81. Malden: Wiley-Blackwell.  https://doi.org/10.1002/9781118257227.ch4 Hernández-Campoy, J. M. and T. García-Vidal. 2018a. Persona management and identity projection in English Medieval Society: Evidence from John Paston II. Journal of Historical Socio­ linguistics 4 (1). 1–31.  https://doi.org/10.1515/jhsl-2016-0027 Hernández-Campoy, J. M. and T. García-Vidal. 2018b. Style-shifting and accommodative competence in Late Middle English written correspondence: Putting audience design to the test of time. Folia Linguistica Historica 39, 383–420. Hernández-Campoy, J. M., J. C. Conde-Silvestre and T. García-Vidal. 2019. Tracing patterns of intra-speaker variation in historical corpora of English correspondence: Data from HiStylVar Project. Studia Anglica Posnaniensia, in press. Hernández-Campoy, J. M. Forthcoming. Corpus-based individual lifespan change in Late Middle English. In K. V. Beaman and I. Buchstaller (eds.), Language variation and language change across the lifespan: Theoretical and empirical perspectives from panel studies. New York: Routledge. Labov, William. 1966. The social stratification of English in New York City. Washington D.C.: Center for Applied Linguistics. Labov, William. 1994. Principles of linguistic change I: Internal factors. Oxford: Blackwell. Maegaard, Marie, Torben Jensen, T. Kristiansen and J. N. Jørgensen. 2013. Diffusion of language change: Accommodation to a moving target. Journal of Sociolinguistics 17 (1). 3–36. https://doi.org/10.1111/josl.12002

116 Juan Manuel Hernández-Campoy

Martínez-Martín, F. 1983. Fonética y sociolingüística en la ciudad de Burgos. Madrid: CSIC. Monroy, R. and J. M. Hernández-Campoy. 2015. Illustrations of the IPA: Murcian Spanish. Journal of the International Phonetic Association 45 (2). 229–240. https://doi.org/10.1017/S0025100314000231 Nevala, Minna and Minna Palander-Collin 2005. Letters and letter writing: Introduction. European Journal of English Studies 9 (1). 1–7.  https://doi.org/10.1080/13825570500067903 Nevalainen, Terttu and Helena Raumolin-Brunberg. (eds.) 1996. Sociolinguistics and language history. Studies based on the Corpus of Early English Correspondence. Amsterdam: Rodopi. Nevalainen, Terttu and Helena Raumolin-Brunberg. 2003. Historical sociolinguistics. Language change in Tudor and Stuart England. London: Longman Pearson Education. Nevalainen, Terttu and S. K. Tanskanen. (eds.) 2007. Letter writing. Amsterdam-Philadelphia: John Benjamins Publishing.  https://doi.org/10.1075/bct.1 Nordberg, Bengt and Eva Sundgren. 1998. On observing real-time language change: A Swedish case study. Uppsala: Uppsala University. Pallander-Collin, Minna. 2010. Correspondence. In A. H. Jucker and I. Taavitsainen (eds.), The handbook of historical pragmatics, 677–703. Berlin-New York: de Gruyter. Penny, Ralph. 1991. A history of the Spanish language. Cambridge: Cambridge University Press Penny, Ralph. 2000. Variation and change in Spanish. Cambridge: Cambridge University Press. https://doi.org/10.1017/CBO9781139164566 Robson, Colin. 1993/2002. Real world research. Oxford: Blackwell. Sankoff, Gillian. 2006a. Cross-sectional and longitudinal studies. In U. Ammon, N. Dittmar, K. J. Mattheier and P. Trudgill (eds.), Sociolinguistics: An international handbook of the science of language and society (Vol. 2), 1003–1013. Berlin: de Gruyter. Sankoff, Gillian. 2006b. Age: Apparent time and real time. In K. Brown (ed.), Elsevier encyclopedia of language and linguistics, 573–579. Oxford: Elsevier. https://doi.org/10.1016/B0-08-044854-2/01479-6 Sankoff, Gillian. 2006c. Age grading in retrograde movement: The inflected future in Montréal French. University of Pennsylvania Working Papers in Linguistics 12 (2). 1–14. Sankoff, Gillian. 2013. Longitudinal studies. In Robert Bayley, Richard Cameron and Ceil Lucas (eds.), The Oxford handbook of sociolinguistics, 261–279. Oxford: Oxford University Press. Schäfer, Ursula. 1996. The Late Middle English Paston letters: A grammatical case in point for reconsidering philological methodologies. In Jürgen Klein and Dirk Vanderbeke (eds.), Anglistentag 1995 Greifswald, 313–323. Tübingen: Niemeyer. https://doi.org/10.1515/9783111714141-028 Schneider, Edgar. 2002. Investigating variation and change in written documents. In J. K. Chambers, P. J. Trudgill and N. Schilling-Estes (eds.), The handbook of language variation and change, 67–96. Oxford: Blackwell. Stenroos, Merja. 2004. Regional dialects and spelling conventions in Late Middle English. Searches for (th) in LALME data. In M. Dossena and R. Lass (eds.), Methods and data in English historical dialectology, 257–285. Bern: Peter Lang. Stenroos, Merja. 2006. A Middle English mess of fricative spellings: Reflections on thorn, yogh and their rivals. In M. Krygier and L. Sikorska (eds.), To make his English sweete upon his tonge, 9–35. Frankfurt: Peter Lang. Tillery J. and G. Bailey. 2003. Approaches to real time in dialectology and sociolinguistics. World Englishes 22. 351–365.  https://doi.org/10.1111/j.1467-971X.2003.00305.x Trudgill, Peter. 1988. Norwich revisited: Recent linguistic changes in an English urban dialect. English World-Wide 9. 33–49.  https://doi.org/10.1075/eww.9.1.03tru



Chapter 6.  Virtual sociolinguistics 117

Van de Velde, Hans, Marinel Gerritsen and Roeland Van Hout. 1996. The devoicing of fricatives in standard Dutch: A real-time study based on radio recordings. Language Variation and Change 8. 149–175.  https://doi.org/10.1017/S0954394500001125 Van de Velde, Hans, Roeland Van Hout and Marinel Gerritsen. 1997. Watching Dutch change: A real time study of variation and change in standard Dutch pronunciation. Journal of Socio­ linguistics 1 (3). 361–391.  https://doi.org/10.1111/1467-9481.00021 Wagner, Suzanne. 2012. Age grading in sociolinguistic theory. Language and Linguistics Compass 6 (6). 371–382.  https://doi.org/10.1002/lnc3.343 Wagner, Suzanne and Isabelle Buchstaller. (eds.) 2018. Panel studies of variation and change. New York: Routledge. Zamora-Vicente, Alonso. 1960/1989. Dialectología española. Madrid: Gredos. Zhang, Qing. (2012). ‘Carry shopping through to the end’: Linguistic innovation in a Chinese television program. In J. M. Hernández-Campoy and J. A. Cutillas-Espinosa (eds.), Style-shifting in public: New perspectives on stylistic variation, 205–224. Amsterdam-Philadelphia: John Benjamins Publishing.  https://doi.org/10.1075/silv.9.11zha

Chapter 7

ASPA Tools or how to measure foreign-accentedness and intelligibility in an objective manner María Ángeles Jurado-Bravo and Gitte Kristiansen

Universidad Complutense de Madrid, Spain / Universidad Rey Juan Carlos I, Spain

Dialectometry has traditionally been examined from the perspective of the dialects and accents of a single language or related languages. However, foreign accents still remain understudied within this paradigm (see section 2 and Wieling et al. 2014 for an exception). In this chapter we present ASPA Tools (Accented-Speech Phonetic Alignment), a web application which measures phonetic distances between foreign-accented speech and a standard pronunciation. Unlike similar instruments (Visual DialectoMetry or Gabmap), ASPA Tools measures the degree of intelligibility of non-native English speech in relation to English as a Lingua Franca (ELF). We use Multiple Sequence Alignment (MSA) and two variations of the Levenshtein Distance (LD) to examine the pronunciation of a group of speakers and compare it to a given ELF standard. Furthermore, ASPA Tools objectively measures the prototypical pronunciation of the group, which allows researchers to analyse the most salient deviations from the ELF standard. Keywords: Lectometry, intelligibility, foreign accents, ELF, ASPA Tools, prototipicality

1. Introduction1 As pointed out in Jurado-Bravo (2018: 5), dialectometry has traditionally been examined from the perspective of the dialects and accents of a single language (Heeringa 2004; Nerbonne and Heeringa 2001; Wieling 2012) and only more 1. Financial support from the Spanish Ministry of Education (FPU grant FPU13/02700) is gratefully acknowledged. https://doi.org/10.1075/silv.22.07jur © 2019 John Benjamins Publishing Company

120 María Ángeles Jurado-Bravo and Gitte Kristiansen

recently from the perspective of the relation between related languages (Beijering, Gooskens, and Heeringa 2008; Gooskens and Heeringa 2004; Gooskens, Heeringa, and Beijering 2008; Heeringa et al. 2006). However, the measurement of phonetic distances of foreign-accented speech with regards to the target language has rarely been addressed until recently (Bloem et al. 2016; Wieling et al. 2014). Instead, foreign accents have generally been analysed following qualitative methodologies (Pilus 2013; Scales et al. 2006). This relative lack of research which implements lectometric methods might in part be due to the software which is currently available to perform such analyses (see section 2 for a description of the available tools). Dialectometry nowadays is an inherently computerised area of investigation, yet the specific design of existent specialised computer software developed to study dialect geography paradoxically hinders the possibility to study foreign accents using dialectometric methodologies. It is for this reason that a new tool specifically designed to measure phonetic distances of foreign-accented speech was deemed necessary. While other instruments analyse dialects in terms of their relative distances or similarities, the application we present in this chapter, and which we have called ASPA Tools (ASPA stands for Accented-Speech Phonetic Alignment), aims at measuring the degree of intelligibility of a speaker or a group of speakers. The reference point is not a specific native English variety like RP or General American, but English as a Lingua Franca (ELF). We define ELF as “any use of English among speakers of different first languages for whom English is the communicative medium of choice, and often the only option” (Seidlhofer 2011, 7). Furthermore, due to the great number of extralinguistic factors affecting the pronunciation of non-native speakers (NNS) of a language, foreign-accented speech presents intrinsic variability within groups of speakers sharing the same linguistic background. Thus, we believe that Prototype Theory (see section 4 below) may contribute substantially to a better understanding of the mechanisms of foreign-accented speech within speech communities. The rest of the chapter is structured as follows. Section 2 reviews the characteristics of existent dialectometric tools. Section 3 describes ELF and Jenkins’s (2000) Lingua Franca Core (LFC), the models that constitute the theoretical foundations for the development of ASPA Tools. Section 4 focuses on Prototype Theory. Section 5 summarises the technical characteristics of ASPA Tools. Finally, section 6 draws a number of conclusions and discusses future implementations of the software. We are thankful to Christian Van de Walle for his excellent programming skills. ASPA Tools would not be possible without him. We would also like to thank Benedikt Szmrecsanyi (KU Leuven), Charlotte Gooskens (University of Groningen), Martijn Wieling (University of Groningen) and Anja Schüppert (University of Groningen) for their valuable feedback and comments in order to improve the application.



Chapter 7.  ASPA Tools or how to measure foreign-accentedness and intelligibility 121

2. Existent dialectometric tools To date, a limited number of dialectometric tools have been developed. These incorporate a wide range of analyses and statistical tests which allow the researcher to explore dialectal data in depth and visualise the results in attractive and comprehensible manners. However, and crucially, none of them seems to be entirely suitable for the type of analysis we want to perform, namely, the calculation of the degree of intelligibility of NNS of English. a. VisualDialectoMetry (VDM), developed at the University of Salzburg, deals with the analysis of linguistic atlas data. It “supports three groups of tasks: the management of classified atlas data, the calculation of similarity matrices and the visualisation of the structure found in the data using different map types” (Haimerl 2006, 438). b. Gabmap (Nerbonne et  al. 2011), an online application developed at the University of Groningen, allows the researcher to work with string data (i.e. phonetic transcriptions), numeric data, categorical data and difference data. Among the great number of tests which could be performed to analyse the data, Gabmap creates distribution maps, calculates linguistic distances using VC-sensitive Levenshtein algorithm (Heeringa 2004), and performs multidimensional scaling and clustering analyses. See Nerbonne et al. (2011) and Leinonen et al. (2016) for a more detailed description of this online application. c. DiaTech (Aurrekoetxea et  al. 2016) is a more recent development at the Universidad del País Vasco (UPV/EHU). An interesting implementation of this application is the possibility to account for multiple responses for the same question and location. “Many questions elicit multiple responses, indicating that the informant would recognise all the responses as dialectally appropriate. Multiple responses are even more common if we aggregate responses from all individual informants in a given community” (Nerbonne and Kleiweg 2003, 349). DiaTech calculates the similarity coefficient of the multiple responses by applying Dice’s coefficient (Aurrekoetxea et al. 2016, 8). Furthermore, this tool offers three types of map representations as well as cluster analysis tools for the presentation of the results. d. ALPHAMALIG (Alonso et al. 2004) applies Multiple Sequence Alignment (MSA) to the data in order to find patterns. Even though MSA has generally been used in bioinformatics, it has also been successfully implemented in linguistic research (Alonso et al. 2004; Prokić, Wieling, and Nerbonne 2009). This application requires the researcher to specify the alphabet and the distance between each character of the alphabet, including gaps.

122 María Ángeles Jurado-Bravo and Gitte Kristiansen

3. English as a Lingua Franca (ELF) and the Lingua Franca Core (LFC) English as a Lingua Franca (ELF) is defined as “any use of English among speakers of different first languages for whom English is the communicative medium of choice, and often the only option” (Seidlhofer 2011, 7). One of the main tenets behind ELF is the intelligibility principle, which “recognises that communication can be remarkably successful when foreign accents are noticeable or even strong” (Levis 2005, 370). Many definitions of the term intelligibility have been provided in research throughout the last decades, but the one acknowledged in this chapter is that by Smith and Nelson (1985, 334), i.e. related to the recognition of words, and therefore focusing on the production of sounds. When Jenkins (2000) realised that deviations in pronunciation accounted for the majority of misunderstandings in ELF communication, she launched the idea of the Lingua Franca Core (LFC), a list of pronunciation features which –ideally speaking– should be mastered by all NNS of English in order to achieve international intelligibility. Jenkins divides English segmental and suprasegmental features into core and non-core items. Core features are those aspects which should be articulated properly, such as “most consonant sounds, appropriate consonant cluster simplification, vowel length distinctions and nuclear stress” (Jenkins 2000, 132). Non-core features, on the other hand, are those where the speaker may deviate from any native standard (e.g. RP, or General American) without causing a breakdown in communication, therefore allowing the speakers to express variation and manifest their linguistic and social identities. Examples of non-core features include vowel quality, weak forms and intonation, among others. Crucially, given the main aims of ASPA Tools, we believe that the establishment of LFC pronunciation as a benchmark for comparison – in spite of the fact that LFC is an abstraction – is preferable to having a native variety act as a standard reference point, for in such a case we would measure the degree of accentedness of the speaker rather than his or her degree of intelligibility. 4. Prototype Theory applied to (foreign) accents Accents and dialects constitute prototype categories (Kristiansen 2003). When we hear a prototypical instantiation of a known (native or foreign) accent we immediately recognise it as an unequivocally good example of the speech variety in question (i.e. as real RP, General American, Cockney, Danish from the Copenhagen area or Spanish from Seville or Barcelona). We may also recognise non-prototypical accents as being from the area in question, but as less good examples of the lect in question.



Chapter 7.  ASPA Tools or how to measure foreign-accentedness and intelligibility 123

According to Prototype Theory (Rosch 1973, 1975a, 1975b; Geeraerts 1989; Taylor 1995) central or prototypical senses may be defined both from a psychological and an analytical perspective. As a psychological notion, the prototype is the first sense, or image, that springs to the native speaker’s mind when a concept is evoked (e.g. a chair or a table for the category of furniture rather than a lamp or an ashtray, just as the nightingale is estimated to be a better example of bird than the penguin or the ostrich). However, it is also possible to determine prototypical meaning analytically by examining which sense is the most central one (i.e. which sense partially overlaps with most other senses in its vicinity, and from which the others seem to derive). For accents to be socially diagnostic (Kristiansen 2003, 2006, 2008), the ability to correlate a stretch of unclassified speech (tokens) with a series of regionally grounded models of pronunciation (types) must gradually develop in our social and linguistic experience. The critical period when such awareness increases significantly for native accents has been established empirically for Spanish children at the age of 7-8 (Kristiansen 2010). When researchers in perceptual dialectology and lectometry deal with regional native and foreign accents, they thus inevitably meet with a range of dialect-internal varieties, some of which will be more typical of the lect in question, while others sound like less good examples, less central and more marginal tokens. Needless to say it is pivotal to be able to measure, for whatever purpose, how central or prototypical a given speech sample is, so as to select the most representative one. ASPA Tools has been developed to deal with foreign accent lectometry and with a proper, objective selection of prototypical speech samples. 5. Technical characteristics of ASPA Tools ASPA Tools presents a user-friendly and intuitive interface, allowing researchers to easily upload and analyse data without requiring them to have previous programming skills. 5.1

Input/Output data

ASPA Tools works with plain text files saved in Unicode format so that the application can interpret the phonetic symbols. ASPA Tools works with the International Phonetic Alphabet (IPA), in a wish to use an alphabet which is common to all languages and is easy to learn and use. The only difference in the symbols used with respect to the IPA chart is the symbol /ʌ/, which makes reference to the central

124 María Ángeles Jurado-Bravo and Gitte Kristiansen

near-open spread vowel instead of the back open-mid spread one. We decided to include this change because English phonology research has traditionally used /ʌ/ to describe this vowel phoneme. Three types of documents are required as input data, namely the orthographic text, the ELF standard transcription to which all files in the corpus are compared, and the corpus files, which include the transcriptions of the speakers under study. As researchers need to create a personal account, projects are saved online for future working sessions, thus saving much time. ASPA Tools works in a single workspace, in which several sections can be shown or hidden to enable the researcher to adjust the workspace to their needs at any moment. We believe this interface saves much time because the user does not have to use several tabs or windows at the same time to access and analyse the results (Figure 1).2

Figure 1.  ASPA Tools workspace

2. To illustrate in adequate manners how ASPA Tools works we have used a small subset of the Speech Accent Archive (Weinberger 2015). All the figures and examples provided are the result of the analysis of these data.



Chapter 7.  ASPA Tools or how to measure foreign-accentedness and intelligibility 125

Output information is mainly provided in tables, which enables the researcher to easily access the results. Such tables provide the following information: a. The normalised ELF-based Levenshtein Distance (LD) (see below for a description of this algorithm) between the ELF standard pronunciation and each of the speakers of the group (Figure 2). This information enables the researcher to rank speakers according to their distance with respect to the ELF standard. b. The normalised ELF-based LD between the prototypical pronunciation of a group and each of the speakers of the group (Figure 2). This information enables the researcher to rank speakers according to their relative distance with respect to the prototype.

Figure 2.  Corpus section. The second column shows the normalised ELF-LD between the speakers and the ELF standard. The third column shows the normalised ELF-LD between the speakers and the prototype

c. The normalised ELF-based LD between the ELF standard and the prototypical pronunciation of each of the words in the text (Figure 3). This information allows the researcher to easily spot words which are a potential threat to intelligibility. d. The details of the word alignments, showing the ELF standard pronunciation in the first row, followed by the aligned transcriptions taken from all the corpus files, and the prototypical pronunciation of that word (Figure 4). It also provides information about the frequency of occurrence of each phonetic variant produced for each segment in the word.

126 María Ángeles Jurado-Bravo and Gitte Kristiansen

Figure 3.  Prototypical pronunciation of words and normalised ELF-LD to the ELF standard

Figure 4.  Details of the alignment of a word in the corpus

5.2

Algorithms

At present, the application includes three related algorithms which help to measure the degree of intelligibility of the speakers:



Chapter 7.  ASPA Tools or how to measure foreign-accentedness and intelligibility 127

a. PMI-based (Pointwise Mutual Information) Levenshtein Distance (PMI-LD). Following Wieling (2012), we decided to apply this variation of the LD to align our samples because it allows us to obtain meaningful alignments with a minimum error rate. Taking as the basis the VC-sensitive LD, this algorithm performs several alignment iterations to look for meaningful distances based on the probability of alignment of two segments (in our case, phonetic symbols). These distances are only used for the sake of obtaining meaningful alignments, but not to calculate the final distances which signal the degree of intelligibility of the speakers. b. Multiple Sequence Alignment (MSA). Together with the calculation of PMI-based distances, ASPA Tools applies MSA to the whole corpus, lining up the transcription of a selected word by taking the data from each of the corpus files. MSA is also implemented to calculate the prototypical pronunciation of the words so that the researcher can study the pronunciation features of a certain group of speakers. c. ELF-based Levenshtein Distance algorithm (LD). This algorithm calculates the phonetic distance between a speaker and the selected benchmark (either the ELF standard or the prototypical pronunciation). The variation of the LD we present here is based on ELF and the LFC (see section 2), thus deviations in core items are assigned a large distance (i.e. 1), while the distance applied to deviations in non-core items is zero. As a result, the higher the obtained final distance with respect to the ELF standard, the less intelligible the speaker will be. Similarly, the higher the distance with respect to the prototype, the less prototypical the speaker is calculated to be. Figures 5 and 6 show two examples of the application of this algorithm to real data. Figure 5 shows that the speaker pronounces this word in a barely intelligible manner according to the LFC; that is, there is no aspiration of the voiceless plosive, the vowel is too short and the final consonant sound is not accurately pronounced (i.e. the speaker produces a voiceless instead of a voiced consonant). On the other hand, Figure 6 presents the distance between the speaker and the standard of a word which is more standard-like. Therefore, the distance applied is shorter (the only deviation with respect to the standard is the shortening of the vowel).

Figure 5.  Example of the pairwise string alignment for one word between the ELF standard and one speaker. The turquoise numbers represent the distance between the speaker’s production and the standard

128 María Ángeles Jurado-Bravo and Gitte Kristiansen

Figure 6.  Example of the pairwise string alignment for one word between the ELF standard and one speaker. The turquoise numbers represent the distance between the speaker’s production and the standard

For these reasons we believe that the application constitutes an excellent method for measuring the intelligibility of accented speech with regards to ELF: not every change with respect to the standard affects international intelligibility in the same manner. The custom scores are included in the database, and therefore the researcher does not need to upload this information when creating a new project. Custom scores cannot be modified by users but are always visible during the working session as a reference during the interpretation of the results. At present, only the information related to the Spanish characteristics has been implemented (following the works by Jenkins [2000] and Walker [2010]), but other language-specific LFC sets will be included in future versions of the application. 6. Conclusions and future perspectives We believe that ASPA Tools constitutes a highly useful tool for both dialectometry and ELF research, enabling researchers to investigate varieties of English within a quantitative approach. We see many advantages in the application, from the objective description of intra-group variation to the analysis of pronunciation features which may hinder ELF intelligibility. The fact that ASPA Tools has been designed to focus on foreign-accented speech, and more specifically on the intelligibility of foreign accents, allows the researcher to analyse this type of data using a completely different methodology from the one traditionally used in this area of research. In our opinion, further research is still necessary before – and if – a Lingua Franca Core can be established on less subjective grounds. The distances applied to the different alignments as described in this chapter have been decided on the basis of current research within the field of the LFC – see Jenkins (2000) for a general description of the items included, and Walker (2010) for the adaptation to Spanish speakers of English. Nevertheless, the scores applied in this version of ASPA Tools may easily be modified in accordance with future applications in this area so that intelligibility scores will remain precise, and, crucially, if the LFC theoretical framework undergoes significant changes, as is to be expected, when put to empirical



Chapter 7.  ASPA Tools or how to measure foreign-accentedness and intelligibility 129

testing. Only through solid experimental research implementing intelligibility tests, may claims of the type that in the LFC framework vowel length matters, but vowel quality does not, be objectively verified. In any case, ASPA Tools is prepared to deal with such new parameters.

References Alonso, Laura, Irene Castellon, Jordi Escribano, Xavier Messeguer, and Lluis Padro. 2004. “Multiple Sequence Alignment for Characterizing the Linear Structure of Revision.” Pro­ ceedings of the 4th International Conference on Language Resources and Evaluation: 403–6. Lisbon (Portugal). Retrieved from: https://goo.gl/C1XNia [Accessed 16/08/2018] Aurrekoetxea, Gotzon, Gotzon Santander, Iker Usobiaga, and Aitor Iglesias. 2016. “Diatech: Tool for Making Dialectometry Easier.” Dialectologia: Revista Electrònica 17: 1–22. Beijering, Karin, Charlotte Gooskens, and Wilbert Heeringa. 2008. “Predicting Intelligibility and Perceived Linguistic Distances by Means of the Levenshtein Algorithm.” Linguistics in the Netherlands 15: 13–24. Bloem, Jelke, Anna Mészáros, Martijn Benjamin Wieling, and John Nerbonne. 2016. “Automatically Identifying Characteristic Features of Non-Native English Accents.” In The Future of Dialects, ed. by Marie-Hélène Côté,, Remco Knooihuizen, and John Nerbonne, 155–173. Berlin: Language Science Press. Geeraerts, Dirk. 1989. Prospects and Problems of Prototype Theory. Linguistics 27 (4): 587–612. https://doi.org/10.1515/ling.1989.27.4.587 Gooskens, Charlotte, and Wilbert Heeringa. 2004. “Perceptive Evaluation of Levenshtein Dialect Distance Measurements Using Norwegian Dialect Data.” Language Variation and Change 16 (03): 189–207.  https://doi.org/10.1017/S0954394504163023 Gooskens, Charlotte, Wilbert Heeringa, and Karin Beijering. 2008. “Phonetic and Lexical Predictors of Intelligibility.” International Journal of Humanities and Arts Computing 2 (1–2): 63–81. https://doi.org/10.3366/E1753854809000317

Haimerl, Edgar. 2006. “Database Design and Technical Solutions for the Management, Calculation, and Visualization of Dialect Mass Data.” Literary and Linguistic Computing 21 (4): 437–444.  https://doi.org/10.1093/llc/fql037 Heeringa, Wilbert. 2004. Measuring Dialect Pronunciation Differences Using Levenshtein Distance. Groningen Dissertations in Linguistics 46. Heeringa, Wilbert, Peter Kleiweg, Charlotte Gooskens, and John Nerbonne. 2006. “Evaluation of String Distance Algorithms for Dialectology.” LD´06 Proceedings of the Workshop on Linguistic Distances: 51–62. Sydney (Australia): Association for Computational Linguistics. Retrieved from: https://goo.gl/ErMoTi [Accessed 26/07/2018] https://doi.org/10.3115/1641976.1641984 Jenkins, Jennifer. 2000. The Phonology of English as an International Language. Oxford Applied Linguistics. Oxford: Oxford University Press. Jurado-Bravo, Maria Angeles. 2018. The Intelligibility of Spanish Regional Accents in L2 English after Tailored Lingua Franca Core Instruction: a Critical Empirical Analysis. Unpublished PhD dissertation. Faculty of Philology, Universidad Complutense de Madrid.

130 María Ángeles Jurado-Bravo and Gitte Kristiansen

Kristiansen, Gitte. 2003. “How to Do Things with Allophones. Linguistic Stereotypes as Cog­ nitive Reference Points in Social Cognition”. In Cognitive Models and Thought. Ideology, Metaphors and Meanings, ed. by René Dirven and Roslyn M. Frank, 69–120. Berlin/New York: Mouton de Gruyter.  https://doi.org/10.1515/9783110892901.69 Kristiansen, Gitte. 2006. “Towards a Usage-Based Cognitive Phonology”. International Journal of English Studies 6 (2): 107–140. Kristiansen, Gitte. 2008. “Style-Shifting and Shifting Styles: A Socio-Cognitive Approach to Lectal Variation”. In Cognitive Sociolinguistics. Language Variation, Cultural Models, Social Systems, ed. by Gitte Kristiansen and René Dirven, 45–88. Berlin, New York: Mouton de Gruyter. https://doi.org/10.1515/9783110199154.1.45 Kristiansen, Gitte. 2010. “Lectal Acquisition and Linguistic Stereotype Formation”. In Advances in Cognitive Sociolinguistics, ed. by Dirk Geeraerts, Gitte Kristiansen and Yves Peirsman, 225–263. Berlin, New York: Mouton de Gruyter.  https://doi.org/10.1515/9783110226461.223 Leinonen, Therese, Çağrı Çöltekin, and John Nerbonne. 2016. “Using Gabmap.” Lingua, Lin­ guistic Research in the CLARIN Infrastructure: 178: 71–83. Levis, John M. 2005. “Changing Contexts and Shifting Paradigms in Pronunciation Teaching.” TESOL Quarterly 39 (3): 369–77.  https://doi.org/10.2307/3588485 Nerbonne, John, Rinke Colen, Charlotte Gooskens, Peter Kleiweg, and Therese Leinonen. 2011. “Gabmap – a Web Application for Dialectology.” Dialectologia: Revista Electrònica: 65–89. Nerbonne, John, and Wilbert Heeringa. 2001. “Computational Comparison and Classification of Dialects.” Dialectologia et Geolinguistica 9: 69–83. Nerbonne, John, and Peter Kleiweg. 2003. “Lexical Distance in LAMSAS.” Computers and the Humanities 37 (3): 339–57.  https://doi.org/10.1023/A:1025042402655 Pilus, Zahariah. 2013. “Exploring ESL Learners’ Attitudes Towards English Accents.” World Ap­ plied Sciences Journal 21: 143–152. Prokić, Jelena, Martijn Benjamin Wieling, and John Nerbonne. 2009. “Multiple Sequence Alignments in Linguistics.” Proceedings of the EACL 2009 Workshop on Language Technology and Resources for Cultural Heritage, Social Sciences, Humanities, and Education, 18–25. Rosch, Eleanor. 1973. “Natural Categories”. Cognitive Psychology 4 (3): 328–350. https://doi.org/10.1016/0010-0285(73)90017-0 Rosch, Eleanor. 1975a. “Cognitive Reference Points”. Cognitive Psychology 7 (4): 532–47. https://doi.org/10.1016/0010-0285(75)90021-3 Rosch, Eleanor. 1975b. “Cognitive Representation of Semantic Categories,” Journal of Experimental Psychology 104 (3): 192–233.  https://doi.org/10.1037/0096-3445.104.3.192 Scales, Julie, Ann Wennerstrom, Dara Richard, and Su Hui Wu. 2006. “Language Learners’ Perceptions of Accent.” Tesol Quarterly 40 (4): 715–738.  https://doi.org/10.2307/40264305 Seidlhofer, Barbara. 2011. Understanding English as a Lingua Franca. Oxford: OUP Oxford. Smith, Larry E., and Cecil L. Nelson. 1985. “International Intelligibility of English: Directions and Resources.” World Englishes 4 (3): 333–342.  https://doi.org/10.1111/j.1467-971X.1985.tb00423.x Taylor, John R. 1995. Linguistic Categorisation: Prototypes in Linguistic Theory. Oxford: Clarendon Press. Walker, Robin. 2010. Teaching the Pronunciation of English as a Lingua Franca. Oxford Hand­ books for Language Teachers. Oxford University Press. Weinberger, Steven. 2015. Speech Accent Archive. George Mason University. Retrieved from http:// accent.gmu.edu [Accessed 10/08/2018]



Chapter 7.  ASPA Tools or how to measure foreign-accentedness and intelligibility 131

Wieling, Martijn Benjamin. 2012. A Quantitative Approach to Social and Geographical Dialect Variation. University Library Groningen: Groningen. Wieling, Martijn Benjamin, Jelke Bloem, Kaitlin Mignella, Mona Timmermeister, and John Nerbonne. 2014. “Measuring Foreign Accent Strength in English. Validating Levenshtein Dis­tance as a Measure.” The Mind Research Repository (Beta) 0 (1):1–12.

Chapter 8

Vowel harmony patterns in Greek dialectal child speech Ioanna Kappa and Marina Tzakosta University of Crete

In this study we test the claim that Vowel Harmony (VH) is universally motivated in child speech and facilitates phonological development in non-harmonic languages (Cohen 2012). We analyse a corpus of Greek L1 developmental data from two children raised in a dialectal environment on the island of Crete and exposed to Standard and dialectal Greek, two varieties with a non-harmonic grammar. The data support the claim that not only does VH seem to be universal in nature but it may also affect the order of vowel acquisition. VH is phonologically conditioned in Greek dialectal child speech and determined by prosodic and positional prominence effects, i.e. stress, directionality and sonority. These effects are evident in the distinct developmental paths adopted by different children (inter-language) or in the speech of one child (intra-language). Keywords: vowel harmony, non-harmonic grammars, dialectal speech, interand intra-child variation, universal morphophonological processes

1. Introduction1 Vowel harmony (henceforth VH) traditionally refers to the assimilatory process applying between non-adjacent vowels. In current phonological theory VH is considered to be a process demanding agreement of specific vocalic feature(s), such as height, roundness, backness, advanced or retracted tongue root (ATR or RTR), within certain grammatical domains (Gafos and Dye 2011: 2164). VH has been widely attested in adult speech in a variety of languages/language families, for example, Turkic languages, Akan, several Bantu languages, Finnish, Hungarian, Iberian languages, among others, resulting in numerous descriptive and theoretical studies

1. The authors would like to thank the anonymous reviewers for their valuable comments and suggestions. All errors are our own. https://doi.org/10.1075/silv.22.08kap © 2019 John Benjamins Publishing Company

134 Ioanna Kappa and Marina Tzakosta

on the attested VH patterns (cf. Anderson 1980; Archangeli and Pulleyblank 2002; Goldsmith 1985; Hayes et al. 2009; Hualde 1989; Walker 2005). Regarding the acquisition of VH, there are well-documented studies for languages with active VH grammars like Finnish (cf. Leiwo et al. 2006) and Turkish (cf. Altan 2007; Altan et al. 2016; Hohenberger et al. 2016, 2017), which display early mastery of language specific VH patterns. There is also experimental evidence that in a harmonic language like Turkish, 6–10 month-old infants display sensitivity to back VH in stem-suffix sequences (cf. Altan et al. 2016; Hohenberger et al. 2016). In a comparative study analysing speech directed to children who acquire Turkish and Hungarian as well as non-harmonic languages such as Farsi and Polish, Ketrez (2014) claims that harmonic languages, like Turkish and Hungarian, provide children with harmony cues which, along with other cues, may be potentially used for word segmentation; however, this is not the case in non-harmonic Farsi and Polish. On the contrary, Mintz et al. (2018) suggest that infants may have a predisposition to use VH as a cue also in a non-harmonic language. Mintz et al. (2018) conducted an experimental study with 7-month-old infants with English as ambient language and the head-turn experiments showed that infants can detect VH patterns and use them as a cue for word segmentation. Lack of VH in child speech is attributed to the fact that VH is morphologically conditioned and morphology is not fully mastered by young learners (cf. Goad 1997, 2001). The emergence of VH in the speech of children who acquire non-harmonic languages like Greek, is actually understudied. In a study on the emergence of VH in child Hebrew, a language with a non-harmonic grammar, Cohen (2012) claims that VH is universally motivated and facilitates phonological development. For the purposes of his study, Cohen analysed naturalistic longitudinal data of two children who acquire Hebrew (child SR, age range 1;02.00–1;07.09 and child RM, age range 1;03.27–2;00.09). He reported a considerable amount of VH in the children’s productions in the initial stage of acquisition, implying that speakers have a universal predisposition for such VH patterns. In his data, three VH patterns were attested. Specifically, VH in the Hebrew child data was motivated by (1) stress, (2) directionality and (3) the sonority/(un)markedness of vowels, as shown in the hierarchy [a > u > i > e > o], with [a] being the most sonorous/ least marked vowel. The two children exhibited distinct developmental paths in the course of vocalic acquisition. SR showed preference for segmental considerations, i.e. vocalic sonority interacted with VH directionality, namely right-to-left in (1a) and left-to-right in (1b). RM showed preference for prosodic considerations, i.e. stress interacted with directionality; stressed vowels of the rightmost syllables, although less sonorous, triggered VH (1c). Both children gradually modified their grammars in order to reach the adult non-harmonic grammar.

Chapter 8.  Vowel harmony patterns in Greek dialectal child speech 135



(1)   (a) (b) (c)

Target

Child

Gloss

Child: Age

όpa túki bakbúk

hápa kúku pupú

‘upsy daisy’ ‘parrot’ ‘bottle’

SR: 01;03.14–1;04.17 SR: 01;03.14–1;04.17 RM: 1;06.05–1;08.01

2. Aims and method of the present study The paper addresses two main questions: (a) what is the domain of application for VH? (b) what drives harmony in the initial stage of acquisition and which factors determine the attested variation in the instances of VH? In other words, the aims of the present study are, first, to challenge and explore the idea of the universal nature of VH as a process facilitating phonological development (cf. Cohen 2012), and, second, to investigate whether VH facilitates or governs vocalic acquisition in Greek, which is also a non-harmonic language. If VH is proved to appear systematically in Greek child speech, we are also interested in investigating the domain of application for VH, the factors which drive harmony and the attested VH patterns. Given the findings which stem from the Hebrew data, the assumptions we make for Greek are the following: a. If VH is active in Hebrew, a non-harmonic language (Cohen 2012), then VH patterns are expected to appear in other non-VH languages. b. If VH appears in Greek child speech, then it is expected to facilitate vocalic acquisition. To test our hypotheses, we draw on Greek L1 naturalistic developmental data from two (2) children who acquire dialectal Greek as a mother language (L1) and are raised in semi-dialectal environments in Crete, specifically, the children are exposed both to Standard Greek and to dialectal Greek, two non-harmonic varieties. The children were recorded on a weekly basis and their age ranged between 1;03–2;09 years (15–33 months). Both children are girls (Child–1/ CHR, age range: 1;05–2;09 and Child–2/ MAR, age range: 1;03–2;07 years). The two girls’ data are quantitatively and qualitatively analysed. The quantitative analysis refers to VH types, i.e. harmonic categories, and not VH tokens, i.e. the exact number of vowel harmony instances emerging in the data; therefore, we provide a rough quantitative analysis in order to propose a formal analysis of vowel harmony patterns.

136 Ioanna Kappa and Marina Tzakosta

3. The data Before discussing the data, it is important to mention that in the dialect of western Crete the ranking of a vowel in the sonority hierarchy determines its role as trigger or target of assimilation within the word and in the context of hiatus resolution. It has been suggested that adult speech in western Cretan dialect satisfies the dialect-specific vowel hierarchy [a>o>u>i, e] proposed by Kappa (2016), which differs from the proposed vowel hierarchies for (adult) Standard Greek in the ranking of the front vowels [i] and [e], namely [o>a>u>i>e] (Hadzidakis 1905; Kaisse 1985) and [a>o>u>i>e] (Malikouti-Drachman and Drachman 1992) as well as from other dialectal vowel hierarchies, like [a>o,u>i>e] in the dialect of Karpathos (cf. Revithiadou et al. 2006). In the dialect of western Crete, both front vowels are equally ranked (both prevail if they appear in a hiatus context), as the less sonorous ones. In this paper we also aim to explore if the children in the present study satisfy the dialect-specific vowel hierarchy or if they exhibit their own distinct vowel hierarchy. Table 1 below signals the total rate of the attested VH patterns during the intermediate developmental phase investigated in this paper.2 It becomes obvious that VH appears rather systematically for both CHR and MAR (18.14% and 11.44% respectively). Table 1.  Rates of VH patterns  

CHR

MAR

All production types VH types %

226  41 18.14

166  19 11.44

Table 2 presents the starting and ending points of the attested VH patterns for CHR and MAR. The findings in this table are summarised as follows: first, VH appears early for both children (CHR–01;05, MAR–01;03) and is attested even in advanced developmental phases (CHR–02;09, MAR–02;02). Second, there are certain principles governing VH, specifically, stress, directionality and sonority, which also drive other aspects of phonological development, like segmental and syllabic structure (cf. Tzakosta 2016). Third, as in other studies on Greek, stress seems to be the major factor which determines the emergence of repair strategies (cf. Tzakosta 2. Following Tzakosta (2004, 2007, 2017), we assume that phonological acquisition is determined by three developmental phases, the initial, the intermediate and the final phase. The intermediate phase can be characterised by sub-stages which display variable patterns and/ or distinct developmental paths.

Chapter 8.  Vowel harmony patterns in Greek dialectal child speech 137



2016). Fourth, stress and directionality seem to be the triggering factors for VH and appear at the same time in Greek child speech. The information provided in Table 2 is further documented in the data in (2)–(5). Table 2.  Starting point of VH  

Stress

Directionality

Sonority/markedness

CHR MAR

01;05–03;02 01;03–01;07

01;05–02;09 01;03–01;07

02;02–02;09 01;11–02;02

The data in (2) indicate that stress is the principal factor that drives VH patterns. In other words, children show faithfulness to the vocalic featural composition of a prosodically and perceptually prominent position, i.e. a stressed syllable (cf. Smith 2002). This is supported by the fact that the stressed vowel is a VH trigger irrespective of its directionality or whether it is more or less sonorous than the target vowel. In (2a), for example, the stressed [i] is less sonorous than the target [a]. In (2b) and (2c), the vocalic triggers also happen to be more sonorous than the targets. However, in (2e) and (2f) stressed [o] and [i], respectively, may also trigger VH.3 (2)   (a) (b) (c) (d) (e) (f) (g) (h) (i)

Target marίna eγó pilótos noná stavrόs fortotís brelók pirúni tenía

Child minίna oγó polótoθ naná vojόθ otitís bolók pulúni inía

Gloss ‘proper name’ ‘I-pers.pron.’ ‘pilot-masc.nom.sg.’ ‘godmother’ ‘cross-masc.nom.sg’ ‘freighter-masc.nom.sg’ ‘keyholder-eut.nom.sg’ ‘fork-neut.nom.sg ‘film-fem.nom.sg

Child: Age CHR: 1;08–1;10 CHR: 1;08–1;10 CHR: 2;02–2;6 MAR: 01;03 MAR:01;06–01;10 MAR: 2:00,24 MAR: 2:01,24 MAR: 2:01,24 MAR: 2;04–2;07

During later stages of the intermediate developmental phase, directionality, i.e. right edge vowel prominence, determines the VH pattern in (3) regardless of the sonority of the trigger or the target vowel. VH may be triggered by an unstressed vowel located at the right edge/word-final syllable resulting in non-iterative regressive assimilation. VH interacts with directionality, i.e. positional faithfulness/ edge prominence.

3. In the children’s productions in (2) and throughout the paper the bold vowel is the trigger and the italicised vowel is the target of VH.

138 Ioanna Kappa and Marina Tzakosta

(3)   (a) (b)

Target fármako átakti

Child fámoko átiti

Gloss ‘medicine’ ‘naughty’

Child: Age CHR: 1;10–2;01 CHR: 1;10–2;01

In advanced acquisitional stages, sonority factors determine the shape of the emergent VH patterns and VH may be triggered by an unstressed, sonorous vowel located at the left edge/word-initial syllable, resulting in non-iterative left-to-right (progressive) assimilation (4a). Sonority may also motivate VH at the expense of directionality, namely a more sonorous, unstressed vowel may affect another unstressed vowel resulting in right-to-left (regressive) assimilation (4b, 4c). (4)   (a) (b) (c)

Target alifí erɣalío peθaméni

Child alafí aɣalíο paθaméni

Gloss ‘cream’ ‘tool’ ‘dead’

Child: Age CHR: 2;02–2;06 CHR: 2;02–2;06 MAR: 2;03–2;06

4. Discussion As already mentioned, the paper investigates (a) the domain of application for VH and (b) the factors that drive harmony in the initial stage of acquisition and determine the attested variation in the instances of VH. Regarding the first question, on the one hand, the Greek data indicate that child VH is not sensitive to morphological structure but it may apply within a stem (5a) or between a stem and an inflectional suffix (5b), (6). (5)  

Target

Child

Gloss

Child: Age

CHR: 2;02–2;6 (a) [pilót-osinfl]ω [polótoθ]ω ‘pilot-masc.nom.sg.’ ‘freighter-masc.nom.sg.’ MAR: 2:00,24 (b) [fortot-ísinfl]ω [otitís]ω (6) [stavr-όsinfl]ω [vojόθ]ω ‘cross-masc.nom.sg’ MAR: 1;06–1;10

We claim that VH may apply only within the harmonic span of two syllables which are located either at the beginning (5a) or at the end (5b) of the word, therefore a non-iterative assimilation always occurs (for the notion span, see McCarthy 2004; for similar findings on the location of the harmonic spans in the Greek dialects, see Revithiadou et al. 2006). In the case of disyllabic children’s productions, the harmonic span coincides with the phonological word (ω) (6). In the children’s productions in (5) and (6) the harmonic span is underlined.

Chapter 8.  Vowel harmony patterns in Greek dialectal child speech 139



On the other hand, VH is driven mainly by faithfulness (Beckman 1998) to the vocalic featural composition of the stressed syllable, which is a prosodically prominent position as well as a perceptually salient one (cf. Smith 2002). In the initial stage of the intermediate phase of acquisition, stress is the major cue for VH, taking precedence over directionality or sonority and VH affects adjacent vowels within the harmonic span (or within the disyllabic ω) resulting in regressive assimilation as in the children’s productions in (2) or in the data in (5) and (6) above. In later stages of acquisition (see data in 3, 4) the attested variation occurs due to the interaction/ competition of various factors, i.e. stress, directionality and sonority/markedness resulting in distinct developmental paths/grammars. Regarding our second research question, which seeks the factors that determine the attested VH patterns, it appears that VH is driven, first, by the preservation of the stressed vowel as a perceptually prominent position (cf. Smith 2002), which, in turn, acts as a VH trigger, and, second, by right-to-left directionality affecting adjacent vowels within ω. In addition, the attested variation occurs due to the interaction and/or competition of various factors, i.e. stress, directionality and sonority/ markedness, which, subsequently, define distinct developmental paths. Schemas 1 and 2 depict the statistical preference (in terms of VH types) for certain VH patterns and VH triggers. More specifically, Schema 1 shows that both CHR and MAR use stress as a major VH trigger. Stress is followed by directionality/ positional faithfulness (PF), namely right edge vowel prominence, and sonority/ markedness (see also Tzakosta 2016). Preferred VH patterns 28

13 10

3

33 Sonority/ markedness

Stress

Directionality/ PF

CHR

28

10

3

MAR

13

3

3

Schema 1.  Preferred patterns of vowel harmony

140 Ioanna Kappa and Marina Tzakosta

Schema 2 illustrates the statistical preference (in terms of VH types) for the selection of the VH triggers. For both CHR (CHR) and MAR (MAR), [o] is the most prevalent CH trigger. Given the analysis, CHR adopts a clear vowel hierarchy, while in the speech of MAR there are two equivalent vocalic pairs, namely, /a/–/e/ and /i/–/u/ (see also Tzakosta 2016). Triggers of VH 13

12 9

7 4

5 4 2

2

Chr

/a/ 12

/e/ 5

/i/ 9

/o/ 13

/u/ 2

Mar

4

4

2

7

2

Schema 2.  Triggers of VH4

Based on the above discussion, it is evident that the emergent VH patterns facilitate the acquisition of Greek vowels. It has previously been claimed that although Greek has a rather limited repertoire of vowels, their acquisition is not perfectly straightforward (Tzakosta 2016). The vocalic hierarchies for CHR and MAR are provided in (7a) and (7b), respectively. Differentiations among children signal distinct developmental paths of vocalic acquisition adopted by different children. (7) a. CHR → o > a > i > e > u b. MAR → o > a, e > i, u

The above vocalic hierarchies (7a, b) adopted by the children mirror the partial disagreement and/or discrepancies of the hierarchies suggested by other researchers for standard Greek and its dialectal variants. The children show a clear preference for non-high vowels, especially for [o] followed by [a] (and [e] for MAR); this preference does not conform to the ranking of the dialect-specific vowel hierarchy [a>o>u>i, e] (Kappa 2016); instead it conforms partially to the vowel hierarchy [o>a>u>i>e] for (adult) Standard Greek (see Hadzidakis 1905; Kaisse 1985). We 4. In the chart, vowels are presented according to their position in the vocal tract, i.e. low, front and back vowels.



Chapter 8.  Vowel harmony patterns in Greek dialectal child speech 141

claim that the phonological properties of the dialectal vowel sonority hierarchy have no effects on VH patterns and the acquisition of vowels, but the children’s specific vocalic hierarchies reflect the above mentioned factors (stress, directionality, sonority) that facilitate the acquisition of the vocalic system of their mother language. To sum up, child Greek VH seems to appear across-the-board during the intermediate developmental phase. Stress seems to be the major cue for VH and it is followed by directionality. Directionality is primarily regressive in nature and applies regardless of whether the trigger or the target of VH is an unmarked vowel or not. Finally, in later stages of the intermediate phase, sonority also determines the shape of the emergent VH patterns. 5. Conclusions In this paper we have explored the idea that VH is universal in nature, a claim initially made by Cohen (2012) for child Hebrew (see also Becker et al. 2011 contra Hayes et al. 2009). Cohen (2012) has claimed that VH is active even in the speech of children who acquire non-harmonic languages; this entails that learners acquire constraints they are not exposed to (contra Vroomen et al. 1998). The Hebrew child data have underlined the fact that VH facilitates phonological acquisition and affects the order of vowel acquisition due to its naturalness (contra Pycha et al. 2003). The Greek data provided further support to the claim made by Cohen (2012), since Greek dialectal child speech exhibits VH patterns which are similar to child Hebrew. More specifically, VH is phonologically conditioned in Greek child speech and is determined by prosodic and positional prominence effects. Stress, directionality and sonority are the principles which govern the emergent VH patterns. However, stress and directionality are the major VH cues, while sonority/ markedness becomes effective in later stages of acquisition. We have also underlined the fact that the domain of VH is a phonological constituent, namely, the phonological word (ω). Finally, VH has been shown to affect the acquisition of the vocalic system of Greek. We suggest that VH may emerge in the speech of children who acquire complex linguistic systems, like Hebrew and Greek. In such systems, phonology and morphology both affect the shape of the emergent forms. We have shown that inter-language VH patterns are governed by phonological properties of the target language while intra-language VH patterns are determined by developmental paths followed by the learner.

142 Ioanna Kappa and Marina Tzakosta

References Altan, Asli. 2007. The acquisition of vowel harmony in Turkish. Paper presented at the OCP4, 2007 University of the Aegean, Rhodes. Altan Aslı, Utku, Kaya and Annette Hohenberger. 2016. Sensitivity of Turkish infants to vowel harmony in Stem- suffix sequences: Preference shift from familiarity to novelty. In Jennifer Scott and Deb Waughtal (eds.), BUCLD 40 Online Proceedings Supplements. Retrieved from http://www.bu.edu/bucld/files/2016/09/BUCLD_Proceedings_2016_01_25_altan1.pdf. Anderson, Stephen R. 1980. Problems and perspectives in the description of vowel harmony. In Robert M. Vago (ed.), Issues in vowel harmony, 1–48. Amsterdam: John Benjamins Publishing. https://doi.org/10.1075/slcs.6.02and Archangeli, Diana and Douglas Pulleyblank. 2002. Kinande vowel harmony: Domains, grounded conditions and one-sided alignment. Phonology 19. 139–188. https://doi.org/10.1017/S095267570200430X Becker, Michael, Nihan Ketrez and Andrew Nevins. 2011. The surfeit of the stimulus: Analytical biases filter lexical statistics in Turkish devoicing neutralization. Language 87 (1). 84–125. https://doi.org/10.1353/lan.2011.0016 Beckman, Jill N. 1998. Positional faithfulness. Doctoral dissertation, University of Massachusetts, Amherst. Cohen, Evan-Gary. 2012. Vowel harmony and universality in Hebrew acquisition. Brill’s Annual of Afroasiatic Languages and Linguistics 4. 7–29.  https://doi.org/10.1163/18776930-00400001 Gafos, Adamantios and Amanda Dye. 2011. Vowel harmony: Transparent and opaque vowels. In Marc van Oostendorp, Colin J. Ewen, Elizabeth Hume and Keren Rice (eds.), The Blackwell companion to phonology, 2164–2189. Malden, MA and Oxford: Wiley-Blackwell. https://doi.org/10.1002/9781444335262.wbctp0091 Goad, Heather. 1997. Consonant harmony in child-language: An optimality-theoretic account. In S. J. Hannahs and Martha Young-Scholten (eds.), Focus on phonological acquisition, 113– 142. Amsterdam: John Benjamins Publishing.  https://doi.org/10.1075/lald.16.06goa Goad, Heather. 2001. Assimilation phenomena and initial constraint ranking in early grammars. In Anna H.–J. Do, Laura Domínguez and Aimee Johansen (eds.), Proceedings of the 25th Annual Boston University Conference on Language Development, 307–318. Somerville, Mass.: Cascadilla Press. Goldsmith, John. 1985. Vowel harmony in Khalkha Mongolian, Yaka, Hungarian and Finnish. Phonology Yearbook 2. 253–75. Cambridge: Cambridge University Press. Hatzidakis, George N. 1905. Medieval and Modern Greek Α΄ (in Greek). Athens. Hayes, Bruce, Péter Siptár, Kie Zuraw and Zsuzsa Londe. 2009. Natural and unnatural constraints in Hungarian vowel harmony. Language 85 (4). 822–863.  https://doi.org/10.1353/lan.0.0169 Hohenberger, Annette, Aslı Altan, Utku Kaya, Özgün Köksal-Tuncer and Enes Avcu. 2016. Sensitivity of Turkish infants to vowel harmony: Preference shift from familiarity to novelty. In Belma Haznedar and Nihan F. Ketrez (eds.), The acquisition of Turkish in childhood, 29–56. Amsterdam: John Benjamins Publishing.  https://doi.org/10.1075/tilar.20.02hon Hohenberger, Annette, Utku Kaya and Aslı Altan. 2017. Discrimination of vowel-harmonic vs vowel-disharmonic words by monolingual Turkish infants in the first year of life. In Maria LaMendola and Jennifer Scott (eds.), Proceedings of the 41st annual Boston University Conference on Language Development, 309–322. Somerville, MA: Cascadilla Press. Hualde, Jose Ignacio. 1989. Autosegmental and metrical spreading in the vowel-harmony systems of Northeastern Spain. Linguistics 27 (5). 773–805.  https://doi.org/10.1515/ling.1989.27.5.773



Chapter 8.  Vowel harmony patterns in Greek dialectal child speech 143

Kaisse, Ellen M. 1985. Connected speech: The interaction of syntax and phonology. New York: Academic Press. Kappa, Ioanna. 2016. Instances of vowel assimilation in the Cretan dialect. In Marina Mattheoudakis and Katerina Nikolaidis (eds.), Selected Papers of the 21st International Symposium on Theoretical and Applied Linguistics (ISTAL 21), 156–165. Thessaloniki: University of Thessaloniki. Ketrez, Nihan F. 2014. Harmonic cues and speech segmentation: A crosslinguistic study on childdirected speech. Journal of Child Language 41 (2). 439–461. https://doi.org/10.1017/S0305000912000724 Leiwo, Matti, Pirjo Kulju and Katsura Aoyama. 2006. The acquisition of Finnish vowel harmony. In Mickael Suominen, Antti Arppe, Anu Airola, Orvokki Heinämäki, Matti Miestamo, Urho Määttä, Jussi Niemi, Kari K. Pitkänen and Kaius Sinnemäki (eds.), A man of measure: Festschrift in honour of Fred Karlsson on his 60th birthday. Special supplement to SKY Journal of Linguistics 19. 149–161. McCarthy, John J. 2004. Headed spans and autosegmental spreading. Unpublished ms. UMass/ Amherst. Retrieved from https://works.bepress.com/john_j_mccarthy/60/ Malikouti-Drachman, Angeliki and Gaberell Drachman. 1992. Greek clitics and lexical phonology. In Wolfgang U. Dressler, Hans Ch. Luschützky, Oskar E. Pfeiffer and John R. Rennisson (eds.), Phonologica 1988. 197–206. Cambridge: Cambridge University Press. Mintz, Toben H., Rachel L. Walker, Ashlee Welday and Celeste Kidd. 2018. Infants’ sensitivity to vowel harmony and its role in segmenting speech. Cognition 171. 95–107. https://doi.org/10.1016/j.cognition.2017.10.020 Pycha, Anne, Pawel Nowak, Eurie Shin and Ryan Shosted. 2003. Phonological rule-learning and its implications for a theory of vowel harmony. In Gina Garding and Mimu Tsujimura (eds.), Proceedings of the 22nd West Coast Conference on Formal Linguistics, 101–114. Somerville: Cascadilla Press. Revithiadou, Anthi, Marc van Oostendorp, Kalomoira Nikolou and Maria-Anna Tiliopoulou. 2006. Vowel harmony in contact‐induced systems: The case of Cappadocian and Silly. In Mark Janse, Brian Joseph and Angela Ralli (eds.), Proceedings of the 2nd International Con­ ference on Modern Greek Dialects and Linguistic Theory, 350–365. Patra: University of Patras. Smith, Jennifer. 2002. Phonological augmentation in prominent positions. Doctoral Dissertation, UMass/Amherst. Tzakosta, Marina. 2004. Multiple parallel grammars in the acquisition of stress in Greek L1. Doctoral Dissertation, LOT Dissertation Series 93. Tzakosta, Marina. 2007. Genetic and environmental effects in L1 phonological acquisition: The case of consonant harmony in Greek child speech. Journal of Greek Linguistics 8. 5–30. https://doi.org/10.1075/jgl.8.04tza Tzakosta, Marina. 2016. A typology of vowel-vowel and consonant-vowel-consonant phenomena in the dialectal variants of Western Crete and its use in education. In Angela Ralli, Nikolaos Koutsoukos and Stavros Bompolas (eds.), Proceedings of the 6th Conference on Modern Greek Dialects and Linguistic Theory, 184–193. Patras: University of Patras. Tzakosta, Marina. 2017. Language genesis: Development in stages or stage-like development? Ms. University of Crete. Vroomen, Jean, Jyrki Tuomainen and Beatrice de Gelder. 1998. The role of word stress and vowel harmony in speech segmentation. Journal of Memory and Language 38 (2). 133–149. https://doi.org/10.1006/jmla.1997.2548 Walker, Rachel. 2005. Weak triggers in vowel harmony. Natural Language and Linguistic Theory 23. 917–989.  https://doi.org/10.1007/s11049-004-4562-z

Chapter 9

Tracking change in social meaning The indexicality of “damped” /i/ in rural Sweden Jenny Nilssoni, Therese Leinonenii and Lena Wenneri

iInstitute

for Language and Folklore, Sweden / iiUniversity of Turku

In this study, we focus on the different connotations of [ɨ] (a variant of the phoneme /i/) in Sweden in order to discuss the possible processes behind the changes in social meaning for a linguistic form.1 We investigate two rural areas (Skärhamn and Edsbyn), where the feature has been part of the traditional dialect, indexing place. In order to investigate how the social meaning of this variant has changed over time we have approached the phenomenon from different angles. Apart from investigating the change in use between the mid–20th century and present time, we have analysed informants’ production, attitudes and specifically perception of [ɨ] today. In Skärhamn [ɨ] still indexes place. In Edsbyn, the social meaning of this variant is more complex. Keywords: indexicality, dialect change, Implicit Association Test, production, perception, interviews, rural areas, Sweden, Swedish dialects

1. Introduction In this study, we focus on the use of the linguistic variant “damped” /i/ ([ɨ], a variant of the phoneme /i/, see e.g. Björsten and Engstrand 1999). This variant has different connotations in different parts of Sweden: in rural areas, it has traditionally indexed (Silverstein 2003) place and locality, while it has indexed the class and gender (higher class female) of speakers in urban Stockholm and Gothenburg (Bruce 2010; Svahn and Nilsson 2014; Öqvist to appear). Previous studies have shown that the urban variant has spread from city centers to less urban areas, and in the process has begun to index urbanity and modernity (see e.g. Svahn and Nilsson 2014). But what happens to this feature in the areas where it has been part of the 1. We are very grateful to Eivind Nessa Torgersen, Jenny Öqvist and to two anonymous reviewers for insightful comments on an earlier draft. We would also like to thank Erika Jönsson for help with conducting the tests as well as Tobias Müller for making the map illustration. https://doi.org/10.1075/silv.22.09nil © 2019 John Benjamins Publishing Company

146 Jenny Nilsson, Therese Leinonen and Lena Wenner

traditional dialect – will it continue to index place and locality, or will its social meaning undergo change? We study the use and perception of “damped” /i/ in two rural villages where it is a traditional dialect variant: Skärhamn in South Bohuslän and Edsbyn in South Hälsingland (see Figure 1). The aims of our study are to (1) investigate the use of “damped” /i/ in the beginning of the 21st century compared with the mid–20th century, and (2) explore how speakers perceive “damped” /i/ today – as rural or urban – and if this is linked to their own attitudes and realisation of /i/. We approach these questions with the help of four data sets: sociolinguistic interviews, production tests, perception tests (Implicit Association Tests) and attitude interviews. By doing so, we wish to answer the above questions as well as to highlight and problematise the processes behind the changes in social meaning for this linguistic form. In the following section, we describe the two investigated areas, Skärhamn and Edsbyn, with special attention to the dialect situation in these communities. In Section 3, the linguistic variant “damped” /i/ is described, followed by an account of the data sets and methods used in this study in Section 4. The results of the study can be found in Section 5, and finally, in Section 6, we summarise and discuss our findings. 2. The locations – Skärhamn and Edsbyn We compare the use, attitudes and perception of “damped” /i/ in two rural villages: Edsbyn in south Hälsingland and Skärhamn in south Bohuslän (see Figure 1). Edsbyn is a small rural, and quite typical, Swedish village. It has approximately 4,000 inhabitants and is situated 35 kilometers from a slightly larger village (Bollnäs with approx. 14,000 inhabitants). There is some daily commuting between Edsbyn and Bollnäs, but Edsbyn has both primary and high schools, some shops, childcare and several industries, thus being a place where you can live your whole life without having to commute elsewhere. This village was investigated by the research project Dialect Change in Rural Sweden.2 Its main objective was to compare processes for linguistic change over 60–70 years in demographically similar locations, and to compare processes for change in these areas (see also Nilsson 2017). The project found that the process for dialect change in Edsbyn followed the same path as in many other locations in Sweden, and Europe (see e.g. Auer 2005): the dialect tended to level towards the standard variety, and there was significant 2. The project ran between 2011 and 2016 and was financed by the Swedish Academy with support from the Knut and Alice Wallenberg Foundations. It also received funds from the Swedish Research Council.

Chapter 9.  Tracking change in social meaning 147



Edsbyn

Stockholm

Skärhamn Gothenburg

Figure 1.  Map of Sweden with Edsbyn and Skärhamn, as well as the two largest cities in Sweden

inter- and intra-individual variation in the community. Younger informants were more levelled than adults, and in those cases where a dialect feature was used infrequently in the recordings from the 1940s, it was usually absent in the recordings from 2013. The dialect spoken in Edsbyn in 2013 consisted to the large part of dialect features spoken within the larger region. Some linguistic features typically found in urban areas were also present, especially in the speech of the adolescents. Skärhamn is in some ways similar to Edsbyn. It is a small rural village (pop. 3.200) and is located 20 kilometres from the larger village Stenungsund (pop. 10.000). However, while there are primary schools in Skärhamn, for (obligatory) high school education adolescents have to commute to Stenungsund. In addition, Skärhamn is not very far (65 km) from Sweden’s second largest city, Gothenburg,

148 Jenny Nilsson, Therese Leinonen and Lena Wenner

and many commute to both Stenungsund and Gothenburg for work. In other words, Skärhamn is small and rural, but its inhabitants in general have more daily contact with urban areas than the Edsbyn population do. The project Dialect levelling in West Sweden investigated dialect change in Skärhamn in 2009 (Svahn and Nilsson 2014).3 The situation was very different from the situation in Edsbyn. In Skärhamn, the inter-individual variation was extensive: there are speakers who use next to no traditional dialect variants at all – rather, there has been a dialect shift towards the Gothenburg variety – alongside very traditional dialect speakers. Further, younger informants tend to mix the Gothenburg variety with more traditional dialect features. Whereas the clear tendency in Edsbyn is leveling towards standard Swedish, there are several processes at work in Skärhamn, possibly influenced by its proximity to Gothenburg. 3. The “damped” /i/ In several Swedish dialects, the vowel /i/ is pronounced in a way that has been described as tjockt ‘thick’, dunkelt ‘dark’, surrande ‘buzzing’ or dämpat “damped” (Engstrand et al 1998: 84). We have chosen to use the term damped. “damped” /i/ is pronounced by a lowered and backed tongue body, indicating a more centralised vowel quality (Westerberg 2016; Schötz et al 2011). It is in general phonetically described as [ɨ]. Most often, this “damped” quality affects the long vowel /i:/, but it also occurs for the short vowel /ɨ/. “Damped” /i/ is subject to regional or social variation, as the vowel occurs in several isolated pockets across Sweden, and seems to have different social functions, or indexicality, in rural versus urban areas (Bruce 2010: 125). In rural areas, on the one hand, “damped” /i/ tends to be a “genuine dialectal feature” (Elert 1995: 45) and can sometimes be perceived as unsophisticated (Bruce 2010: 125). In some urban areas, on the other hand, “damped” /i/ is a “clear prestige marker”, and a sign of high social status (Bruce 2010: 135–136). Originally, it indexed higher class female in Gothenburg and Stockholm, but as it spreads outside these locations it seems to index urbanity and modernity (see e.g. Svahn and Nilsson 2014). In the study of Westerberg (2016) speakers in larger cities have been found to have an auditory and acoustically stronger “damped” /i/ than speakers in smaller locations, and there is a trend towards higher-class females having stronger “damped” /i/.

3. The project ran between 2007 and 2011 and was funded by the Swedish Tercentenary Foundation (Riksbankens Jubileumsfond).

Chapter 9.  Tracking change in social meaning 149



4. Data and methods In order to answer the question of how the use of “damped” /i/ has changed in real time over 70 years in Edsbyn and Skärhamn respectively, we have compared recordings from the mid–20th century with newly recorded data from the projects Dialect levelling in West Sweden as well as Dialect change in rural Sweden (Section 4.1). We have also collected additional data for this study. This dataset (presented in 4.2) consists of a production test (4.2.1), an Implicit Association Test (IAT) (4.2.2) and short attitude interviews (4.4) with another group of informants. 4.1

Data from the project Dialect levelling in West Sweden and the project Dialect change in rural Sweden

The projects Dialect levelling in West Sweden and Dialect change in rural Sweden included sociolinguistic interviews and self-recordings from 2009 and 2013. We have investigated the use of “damped” /i/ in the speech of five adolescents and four adults in Skärhamn as well as six adolescents and twelve adults in Edsbyn. In both locations, we also noted the use of “damped” /i/ in recordings made with informants in the 1940s and 1950s. The older recordings were made with NORM informants born in the latter half of the 19th century. Comparing old NORM recordings with modern sociolinguistic data can be problematic (see e.g. Nilsson 2009) and therefore we have included newly recorded NORM informants (see also Nilsson 2017). This dataset consists of natural speech, with overlaps and, in several of the self-recordings, background noise. Therefore, we have not made any acoustic analyses. As mentioned above, we have noted cases of “damped” /i/, but we present no statistics here. The reason for this is that when /i/ is pronounced without primary stress, it is difficult, at least to our ears, to determine whether it is realised as “damped” or not. It would perhaps be possible to omit all instances of un-stressed /i/; unfortunately, it is not always easy to draw the line between what is stressed and un-stressed in natural speech (Strangert and Heldner 1995), and such a venture risks leading to arbitrary and misleading results. Besides, there are several cases of un-stressed /i/ clearly realized as “damped” (Nilsson 2017; Svahn and Nilsson 2014). 4.2

New data from Edsbyn and Skärhamn

For this study, we have collected new data from Edsbyn (20 informants) and Skärhamn (18 informants). Both men and women participated, and the age distribution was somewhat different at the two sites: the Edsbyn participants were between 18 and 64 years with a median age of 34.5, while the Skärhamn participants were

150 Jenny Nilsson, Therese Leinonen and Lena Wenner

between 22 and 75 years with a median age of 57. These informants all conducted a production test, an IAT test as well as participated in short attitude interviews. 4.2.1 Production test The production test consisted of two parts. In the first part, the informants read a short text written on the computer screen, and in the second part, they read 46 single words appearing on the screen. The text and the words consisted of several tokens of /i/. In addition, the text and words included all Swedish vowels, in order to capture the informants’ whole vowel system. In this study, we present the result of an auditive analysis of this data. 4.2.2 IATs After the participants had conducted the production test, they carried out an Implicit Association test (IAT, Greenwald et al. 1998; Campbell-Kibler 2012). An IAT is an item sorting task, where participants are sorting items into two pairs of categories using a left and a right response key. The categories are combined in two different ways during the test, and the idea is that people should be faster at finding out if the categories assigned to each key can be easily associated with each other. The aim of the IAT tests in our study was to measure to what extent participants associate different pronunciations of /i/ with rurality or urbanity. Audio stimuli were recorded using three different speakers from Edsbyn, Skärhamn and Gothenburg. The stimuli consisted of five different Swedish /Ci:C/ words, and the speakers were instructed to pronounce the words with “damped” /i/ as well as with cardinal /i/. The durations of the stimuli were subsequently manipulated so that all stimuli by the same speaker had identical durations. For each speaker there was, hence, a total of 10 stimuli, i.e. 5 stimulus pairs differing only in the pronunciation of /i/. The stimuli for the categories urban and rural were photos of typical Swedish rural and urban landscapes. The participants in Edsbyn and Skärhamn performed two IATs each: first an IAT with a local speaker (local IAT), and second an IAT with the speaker from Gothenburg (Gothenburg IAT). Table 1 shows the design of the IATs. The order of the tasks was counterbalanced across participants, so that half of the participants had the design shown in Table 1, while the other half had “damped” /i/ on the left side and cardinal /i/ on the right side in blocks 1, 3 and 4 and the other way around in blocks 5–7. The experiment was set up using PsychoPy (Peirce 2007), and participants performed it on a laptop computer with audio-technica headphones. The results consist of one D score (Greenwald, Nosek and Banaji 2003) per IAT test and participant. D scores close to zero mean no IAT effect at all, while positive D scores

Chapter 9.  Tracking change in social meaning 151



Table 1.  The order of tasks in the IATs Block

Trials

Task

Left key

Right key

0a 0b 1 2 3 4 5 6 7

 5  5 20 20 20 40 40 20 40

familiarisation familiarisation discrimination discrimination combined task combined task reversed discrimination reversed combined task reversed combined task

cardinal /i:/ “damped” /i:/ cardinal /i:/ “damped” /i:/ urban rural urban, cardinal /i:/ rural, “damped” /i:/ urban, cardinal /i:/ rural, “damped” /i:/ “damped” /i:/ cardinal /i:/ urban, “damped” /i:/ rural, cardinal /i:/ urban, “damped” /i:/ rural, cardinal /i:/

imply that “damped” /i/ is associated with urbanity and cardinal /i/ with rurality, and negative D scores imply that “damped” /i/ is associated with rurality and cardinal /i/ with urbanity. 4.2.3 Attitude interviews After the production test and IAT test, we conducted short interviews with the informants. The informants were asked questions about what they thought of their own realisation of /i/ as well as their associations and attitudes towards “damped” /i/. The interviews also provided the study with additional production data. 5. Results In 5.1, we present the use of the short and long “damped” /i/ in Edsbyn and Skärhamn in the 21st century compared to the mid–20th century based on the data from the two projects Dialect levelling in West Sweden and Dialect change in rural Sweden. In 5.2, we present the results from the production test, in 5.3 the results from the IAT test and in 5.4 we present the results from the attitude interviews, based on the findings in the dataset collected for this study. 5.1

The use of short and long “damped” /i/ in Edsbyn and Skärhamn

In this section, we present the results from the recordings made within the projects Dialect levelling in West Sweden and Dialect change in rural Sweden. Table 2 show different groups of informants’ use of long and short “damped” /i/ in Skärhamn and Edsbyn.

152 Jenny Nilsson, Therese Leinonen and Lena Wenner

Table 2.  The use of long and short “damped” /i/ in Edsbyn and Skärhamn   Informants Older adults 1940s Older adults 2013 Younger adults 2013 Adolescents 2013

Edsbyn “damped” /i/

 

Skärhamn

“damped”   Informants /i:/

Y

Y

N N

N N/Y*

N

Y/N**

  Older adults 1940s   Older adults 2013   Younger adults 2013   Adolescents 2013

 

“damped” /i/

“damped” /i:/

Y

Y

Y Y

Y Y

Y/N***

Y

Y (yes) and N (no) indicates that all or no informants of that group has at least some tokens of the feature, unless otherwise stated. (*) A few speakers had single tokens. (**) Five out of six had tokens. (***) Four out of five had tokens. In the recordings made in the 1940’s, Edsbyn and Skärhamn informants use both the short and long “damped” /i/. In 2009, Skärhamn speakers still use both the short and long “damped” /i/. By contrast, in the recordings made in 2013 in Edsbyn, none of the older adults uses either short or long “damped” /i/. Among the younger adults, a few have single tokens of the long “damped” /i/, but never the short one. The adolescents frequently use the long “damped” /i/, but never the short variant. 5.2

Results of the production test

Auditive analysis of the data from the production test suggests that the qualitative difference between “damped” /i/ and cardinal /i/ is larger in Skärhamn than in Edsbyn. Determining the exact acoustics of the different variants remains a task for future research. In Edsbyn, nine informants out of 20 use “damped” /i/. These nine informants are all younger than 40, and they only use it for the long variant. The situation in Skärhamn is quite different: 15 informants out of 18 use “damped” /i/, and they (with one exception) use it for both long and short /i/. From the recorded attitude interviews, we could also get a picture of the informants’ use of other dialect features and both locations were characterised by great inter-individual variation.

Chapter 9.  Tracking change in social meaning 153



5.3

Results of the IATs

The boxplots in Figure 2 display the distributions of D scores in both IAT tests at both sites. Most of the participants in Skärhamn have negative D scores, while positive as well as negative D scores are found in Edsbyn. The mainly negative D scores in Skärhamn imply that “damped” /i/ is associated with rurality and cardinal /i/ with urbanity. Two-tailed t-tests were carried out for each test and each site in order to test whether the D scores deviate significantly from zero: Edsbyn local IAT t(19) = –.72, p = .483; Edsbyn Gothenburg IAT t(19) =  – 1.62, p = .123; Skärhamn local IAT t(17) = –3.60, p = .002; Skärhamn Gothenburg IAT t(17) = –3.42, p = .003. The results show a significant IAT effect in both IATs in Skärhamn, but not in Edsbyn. Edsbyn

Skärhamn

1.0

D score

0.5

0.0

–0.5

–1.0 local IAT

Gothenburg IAT

local IAT

Gothenburg IAT

Figure 2.  Distributions of D scores in both IATs in Edsbyn and Skärhamn

As order was counterbalanced across participants in the IAT test, we also tested the effect of order by applying two sample t-tests. Order did not show any significant effects. Since the participants in Edsbyn did not show any significant IAT effect on group level, we also carried out an error analysis, to check whether participants were able to discriminate between the two pronunciations of /i/ above chance level. The correctness was generally above chance level: Edsbyn local IAT M = 65.5%, Edsbyn Gothenburg IAT M = 86.5%, Skärhamn local IAT M = 79.5%, Skärhamn Gothenburg IAT M = 87.7%. More exactly, four informants from Edsbyn had a correctness rate of 50% or less in the local test. All other informants in all other tests had correctness rates above 50%. The participants from Edsbyn performed poorer

154 Jenny Nilsson, Therese Leinonen and Lena Wenner

than Skärhamn participants in local IAT, but as well as Skärhamn participants in Gothenburg IAT. The perceivable difference between “damped” and cardinal /i/ in the Edsbyn stimuli was not as distinct as in the Gothenburg stimuli (the dialect variants were not very “damped”), which might explain the Edsbyn participants’ performance. A closer analysis of the error rates revealed that the Edsbyn participants made more mistakes on cardinal /i/ than on “damped” /i/ in their local test, i.e. they often interpreted cardinal stimuli as “damped”. As mentioned in Section 4, the study did not have a balanced design concerning the age of the participants. Most of the Skärhamn participants were above 40 years old, while the Edsbyn participants had a more scattered age distribution. Therefore, it was not possible to test the effect of age statistically. An inspection of the individual scores still showed that almost all participants above 40 at both sites had negative D scores. The positive D scores from Edsbyn, implying that “damped” /i/ is associated with urbanity, are found mainly among participants younger than 40. 5.4

Results from the attitude interviews

The attitude interviews revealed that informants in Skärhamn were very positive both towards their own dialect and towards “damped” /i/. The reports from the Edsbyn informants were less straightforward: Some reported positive attitudes, but the majority expressed negative opinions towards “damped” /i/ or no opinion at all. At the same time, most informants in Edsbyn reported that they used “damped” /i/, the same as in Skärhamn. 6. Summary and discussion In Skärhamn, the use of both short and long “damped” /i/ is stable and the feature is still actively in use today by all age groups. In the data specifically collected for this study, there is great inter-individual variation among the informants when it comes to the use of dialect – some are very traditional dialect speakers while others are quite standardised. Still, the majority of the informants (15 out of 18) use both long and short “damped” /i/. It is apparent that Skärhamn residents are aware of the use of “damped” /i/. In the perception test, they also easily identify the variant and they clearly associate “damped” /i/ with rurality. The informants report positive attitudes towards both the local dialect and “damped” /i/ and they report that they use the variant themselves. Despite the stability in use over time, there seems to be some change in the social meaning of [ɨ] in Skärhamn. Our attitude interviews indicate that this variant has a more prominent position in today’s speech community than to only index place. The informants repeatedly position themselves towards



Chapter 9.  Tracking change in social meaning 155

“damped” /i/ in a manner that suggests that it indexes authentic local identity and traditional values alongside place (see also Johnstone et al. 2006; Johnstone and Kielsing 2008; Johnstone 2009, 2014). The fact that Skärhamn is located not very far from Sweden’s second largest city (with a different dialect) may affect some of Skärhamn’s citizens to speak more dialect (as is the case in other locations where locals position themselves from visitors linguistically, see e.g. Labov 1963). In Edsbyn, there has been a change in the use of “damped” /i/. As became apparent when comparing newly recorded data to recordings from the mid–20th century, speakers of this area used to realize both long and short /i/ as “damped”, but today only younger residents of Edsbyn use the long “damped” /i/. As previous studies have shown that the long “damped” /i/ indexes urbanity and modernity when it spreads from urban areas to other parts of Sweden, it is possible that the social meaning of “damped” /i/ in Edsbyn is undergoing change, from indexing place towards indexing urban and modern. In our newly recorded data, the Edsbyn informants are in general more levelled towards standard than the informants in Skärhamn, and only nine out of 20 informants use [ɨ]. Interestingly, though, most of these Edsbyn informants report that they use “damped” /i/. The perceived use of this variant obviously differs from the actual use. Further, the results of the IAT tests show that the Edsbyn informants have difficulties identifying “damped” /i/ and that there is a tendency for older speakers to associate “damped” /i/ with rurality, while younger speakers tend to associate it with urbanity. From the attitude interview data, it is clear that there are mixed feelings towards both the traditional dialect and the use of “damped” /i/. Some of the informants were neutral in the discussion, while others reported quite negative attitudes. Apparently, processes for changing social meaning of a specific feature may be different in different locations within the same nation (but cf. Levon, Maegaard and Pharao 2017 for a discussion of a linguistic feature with similar indexicality across languages). The speech communities in Edsbyn and Skärhamn treat “damped” /i/ differently: whereas it seems to index place, local identity and tradition in Skärhamn, the situation in Edsbyn is more complex. The Edsbyn informants claim that “damped” /i/ is a traditional dialect variant and as such index place. However, the IATs tell another story: while there are clear correlations between “damped” /i/ and rural in Skärhamn, there seems to be an ongoing change in Edsbyn. We suggest that the changing use of “damped” /i/ in Edsbyn reflects the complexity of its connotations: on the one hand there are associations that are passed down by previous generations (i.e. local), on the other the urban long /i/ indexing modernity is spreading across Sweden and speakers in Edsbyn might have unconscious connotations that are undergoing change. Finally, we suggest that the differences in the change in use of “damped” /i/ between Edsbyn and Skärhamn also reflect the attitude towards the speakers’ own dialect. Whereas informants in Edsbyn seem

156 Jenny Nilsson, Therese Leinonen and Lena Wenner

to be quite neutral towards their own dialect, the informants in Skärhamn are in general proud of their dialect and the attitude interviews reveal that dialect is important in constructing and maintaining an authentic local identity (see also Røyneland 2005; Johnstone et al. 2006).

References Auer, Peter. 2005. Europe’s sociolinguistic unity, or: A typology of European dialect/standard constellations. In Nicole Delbecque, Johan van der Auwera and Dirk Geeraerts (eds.), Perspectives on variation: Sociolinguistic, historical, comparative, 7–42. (Trends in linguistics. Studies and monographs 163.) Berlin: de Gruyter.  https://doi.org/10.1515/9783110909579.7 Björsten, Sven and Olle Engstrand. 1999. Swedish “damped” /i/ and /y/: Experimental and typological observations. In John J. Ohala (ed.), Proceedings of the 14th International Congress of Phonetic Sciences (ICPhS 99), San Francisco, August 1999, 1957–1960. Berkeley, Calif.: Linguistics department, Univ. of California. Bruce, Gösta. 2010. Vår fonetiska geografi: Om svenskans accenter, melodi och uttal. Lund: Studentlitteratur. Campbell-Kibler, Kathryn. 2012. The implicit association test and sociolinguistic meaning. Lingua 122 (7). 753–763.  https://doi.org/10.1016/j.lingua.2012.01.002 Elert, Claes-Christian. 1995. Allmän och svensk fonetik. (7. omarb. uppl.) Stockholm: Norstedt. Engstrand, Olle, Sven Björsten, Björn Lindblom, Gösta Bruce and Anders Eriksson. 1998. Hur udda är Viby-i? Experimentella och typologiska observationer. Folkmålsstudier 39. 83–95. Greenwald, Anthony G., Debbie E. McGhee and Jordan L. K. Schwartz. 1998. Measuring individual differences in implicit cognition: The Implicit Association Test. Journal of Personality and Social Psychology 74 (6). 1464–1480.  https://doi.org/10.1037/0022-3514.74.6.1464 Greenwald, Anthony G., Brian A. Nosek and Mahzarin R. Banaji. 2003. Understanding and using the Implicit Association Test: I. An improved scoring algorithm. Journal of Personality and Social Psychology 85 (2). 197–216.  https://doi.org/10.1037/0022-3514.85.2.197 Johnstone, Barbara. 2009. Pittsburghese shirts: Commodification and the enregisterment of an urban dialect. American Speech 84 (2). 157–175.  https://doi.org/10.1215/00031283-2009-013 Johnstone, Barbara. 2014. ‘100% Authentic Pittsburgh’: Sociolinguistic authenticity and the linguistics of particularity. In Véronique Lacoste, Jakob Leimgruber and Thiemo Breyer (eds.), Indexing authenticity: Sociolinguistic perspectives, (Linguae and litterae 39). Berlin-München-Boston: DeGruyter.  https://doi.org/10.1515/9783110347012.97 Johnstone, Barbara, Jennifer Andrus and Andrew E. Danielson. 2006. Mobility, indexicality and the enregisterment of Pittsburghese’. Journal of English Linguistics 34 (2). 77–104. https://doi.org/10.1177/0075424206290692 Johnstone, Barbara and Scott F. Kiesling. 2008. Indexicality and experience: Exploring the meanings of /aw/-monophthongisation in Pittsburgh. Journal of Sociolinguistics 12 (1). 5–33.  https://doi.org/10.1111/j.1467-9841.2008.00351.x Labov, William. 1963. The social motivation of a sound change. Word 19. 273–309. https://doi.org/10.1080/00437956.1963.11659799 Levon, Erez, Marie Maegaard and Nicolai Pharao. 2017. Introduction: Tracing the origin of /s/ variation. Linguistics 55 (5). 979–992.  https://doi.org/10.1515/ling-2017-0016



Chapter 9.  Tracking change in social meaning 157

Nilsson, Jenny. 2009. Dialect change? Nordic Journal of Linguistics 32 (2). 207–220. https://doi.org/10.1017/S0332586509990047 Nilsson, Jenny. 2017. Something old, something new: Some processes for dialect change in Sweden. Nordic Journal of Linguistics 40 (3). 1–21.  https://doi.org/10.1017/S0332586517000087 Öqvist, Jenny. To appear. Perspektiv på Stockholmska. Institutet för språk och folkminnen. Peirce, Jon W. 2007. PsychoPy – Psychophysics software in Python. Journal of Neuroscience Methods 162 (1–2). 8–13.  https://doi.org/10.1016/j.jneumeth.2006.11.017 Røyneland, Unn. 2005. Dialektnivellering, ungdom och identitet. Ein komparativ studie av sprakleg variasjon og endring i to tilgrensande dialektområden, Røros og Tynset. (Acta Humaniora 231). Oslo: Det hunamistiske fakultet, Universitetet i Oslo. Schötz, Susanne, Johan Frid and Anders Löfqvist. 2011. Exotic vowels in Swedish – an articulographic and acoustic pilot study of /i:/. In Proceedings of the 17th International Congress of Phonetic Sciences (ICPhS XVII), Hong Kong, 17–21 August 2011.1766–1769. Silverstein, Michael. 2003. Indexical order and the dialectics of sociolinguistic life. Language and Communication 23. 193–229.  https://doi.org/10.1016/S0271-5309(03)00013-2 Strangert, Eva and Mattias Heldner. 1995. Labelling of boundaries and prominences by phonetically experienced and nonexperienced transcribers. Phonum 3. 85–109. Svahn, Margareta and Jenny Nilsson. 2014. Dialektutjämning i Västsverige. Göteborg: Dialekt, -ortnamns- och folkminnesarkivet i Göteborg. Westerberg, Fabienne E. 2016. An auditory, acoustic, articulatory and sociophonetic study of Swedish Viby-i. MPhil(R) thesis. Retrieved from http://theses.gla.ac.uk/7382/, accessed August 01, 2018.

Chapter 10

Slit-t in Dublin English Fergus O’Dwyer

Marino Institute of Education

The paper presents an analysis of speakers in Dublin, Ireland who align themselves with an authoritative interactional identity by realising a slit-t variant. A statistical analysis is first presented and then salient patterns of variation are discussed before proceeding to a qualitative evaluation. The latter highlights the discourse functions clustering around the slit-t occurrences, which indicate that a speaker will adopt a sociopragmatic position, like emphasizing a point. My interpretations characterize the interactions where slit-t tokens are found as epistemically-based, inferring how speakers position themselves and others. The conclusions include a discussion of how ethnographically informed, qualitatively-skewed mixed methods can elucidate the social meaning of linguistic variants. Keywords: discourse analysis, discourse functions of linguistic variants, Dublin English, ethnography, frication, interactional identity, Irish English, slit-t, social meaning, sociopragmatics

1. Introduction: Slit-t in Dublin English Irish English emerged as the result of a complicated history and prolonged language contact but is now an important marker of social identity and solidarity in the modern-day context (e.g. Kirk and Kallen 2011: 270). Previous studies of Dublin English (Hickey 2005; Lonergan 2013) indicate that linguistic variation in Dublin is stratified along geographic, generational and socioeconomic lines. The broad lines of social distinction in Dublin have traditionally been drawn along the north/south division of the Liffey river, which flows through the centre of the capital. Research has shown that terms like northside and southside go beyond physical geography in Dublin and encode class-based distinctions within Dublin (Share 2006: 344). The southside refers primarily to the posh affluent suburbs in the South-East. Northside or skanger has connotation of roughness and disadvantage. Crude stereotypes such as posh or skanger are prevalent, but we do not know yet https://doi.org/10.1075/silv.22.10odw © 2019 John Benjamins Publishing Company

160 Fergus O’Dwyer

how these divisions relate to ways of speaking in Dublin and everyday life in Dublin 15 (this area is the 15th of 22 numbered postcodes in Dublin). Research in Dublin has not, to date, appropriately responded to the call by many sociolinguists (e.g. Eckert 2012) to emphasize social meaning by focusing on the personae and styles adopted by individuals and groups. Recent pragmatics and corpus linguistics work (e.g. Clancy and Vaughan 2012: 237) certainly go a long way towards understanding the dynamics of linguistic variation in Ireland. However, it is still mostly focused on understanding the functions of linguistic items, with little understanding of how – and for which purposes – linguistic variation is performed in modern Dublin and beyond. A refined analysis of the social space in Dublin can provide knowledge about sociolinguistic patterns: this research, for example, can investigate nuances of class-related representations, and how variation is related to practices such as sports or lifestyle choices. The realizations of word-final /t/ in Dublin include a slit-t [t̞ ], an apical alveolar fricative, and glottal realizations ranging between a glottal fricative [h], a glottalised variant [ʔt], glottal stop [ʔ], and something close to deletion. While complete deletion is rare, glottal variants have been related to the rough northside and seem to index tough and aggressive personae. While slit-t is linked to rural varieties of Irish English (Kallen 2013: 54), little has been done in terms of sociolinguistics or pragmatics of the feature. This chapter examines the interactional identities that are linked to a fricated /t/, in order to extend knowledge about how this variant is used in Dublin and beyond. Constructivism in sociolinguistics views aspects of language that are connected with identity as being important, directly functional activities in their own right (Omoniyi 2006: 18), which speakers control and use to their own ends. Individuals position themselves in relation to a set of social images or characteristics, and these attributes are valued differently across various different contexts an individual may operate in. Each positioning taken up entails different kinds of behaviours, practices and manipulation of repertoires including linguistic practices (ibid.). This third wave sociolinguistic study creates an understanding of the salient attributes and ways of being in this specific context, and how they relate to available identities and linguistic variants. My analysis of interactions indicates how speakers project identity on a moment-to-moment basis. I first overview the context where I collected the data, then consider the patterns of variation, before discussing the discourse functions of linguistic variants in question and concluding the paper.



Chapter 10.  Slit-t in Dublin English 161

2. Club Fingal and wider context This study is an ethnography conducted in Club Fingal: a sports club – Gaelic Athletic Association (GAA) – found in the Dublin 15 suburb. Club Fingal is part of a nationwide organization built around local community networks, which primarily focus on the playing of the Gaelic Football and Hurling sports. A GAA player typically expends a large amount of effort and time in a shared enterprise with the same group of players in formative years and beyond. Group identity is also formed when showing support for the local team, with the social centres of clubs designed to bring people together. I chose to conduct a sociolinguistic study in this context as such clubs are fertile ground for identity construction and observation of social interaction. As the interactions I focused on included male-only speakers, the study also provides knowledge about modern male identities. Purposes of the ethnography included to illuminate how individuals in this context position their identity, and to explore the role of language in the process. I now discuss the issues that emerged to be salient in this case: the social dynamics of the locality, the typical linguistic behaviour – types of talk – found in the club, and prominent interactional identities adopted in club communication. The suburb of Dublin 15 is socially stratified, and can be classified socioeconomically as affluent (area 1), above average (area 2), and from below average to disadvantaged (area 3). These areas are associated with two crude Dublin stereotypes: the posh identity is linked to affluence, while the proper Dublin identity is linked to disadvantage and anti-social behaviour (e.g. drug-taking). The posh and proper identities are coordinates for individuals to locate themselves, with a wide range of styles in between. The members of the club typically adopted personae that connotes the demeanour of a community person through lifestyle choices such as socializing in locations that are imbued with a grounded image which is not posh, while also avoiding any links to anti-social behaviour and proper Dublin. I also outlined typical language used in the club, highlighting three types of talk prevalent in the interview data: Dressing Room Talk, Information Talk, and Social Talk. The following describes these briefly: a. The topics of Dressing Room Talk are mostly about games, and what was said in the dressing room before games, for example. This talk is inherently aggressive, and often produced in situations where speakers are trying to command respect or attention regarding issues about the playing of club games. b. In Information Talk speakers inform me, the interviewer, how the club works, for example, and how the speakers are involved in these practices. c. Social Talk is more social in nature, such as casual chatting.

162 Fergus O’Dwyer

An authoritative interactional identity also emerged as salient in Club Fingal. This involves assuming a position of authority or knowledge, as a result of being heavily involved in club activities over a lengthy period, for example. I chose the word-final /t/ variable as data analysis indicates that individuals pronounce this differently in situations where they adopt positions of knowledge or authority, in comparison to situations where talk is more social in nature, for example. The use of slit-t in pragmatic markers (PMs) also emerged to be prominent in this context. PMs in Irish English are often used to foreground an interpersonal approach, by showing other-attentiveness (Amador-Moreno, McCafferty and Vaughan 2015: 12). Features such as now, so or like seem to have particular pragmatic functions: for example, right often lends credibility to the voice of a narrator (Clancy and Vaughan 2012: 226). The structural functions of PMs do not correspond to a lexical meaning, but to a pragmatic meaning of these items: for example, commenting on a topic about a particular utterance (e.g. “But he was a good athlete all right”) allow the hearer to draw conclusions about the speaker’s opinion (Carter and McCarthy 2006: 208). The pragmatic meaning – which indicates a subjective response – of all right in the above example differs from the lexical meaning of all right (‘fairly good’; ‘acceptable’; ‘agreeable’; ‘suitable’ or ‘appropriate’), for example. At the interpersonal level PMs are used to subjectively express attitude, and to achieve intimacy between speakers. PM tokens are often classified according to their pragmatic functions, which is then used as a base for quantitative analysis (e.g. Brinton 2008: 1). 2.1

The data

This paper analyses fourteen interviews I conducted with playing team members between September 2012 and September 2015. The interviews lasted from 55 to 60 minutes and are informal with speech close to that of a casual conversation. Interviews begin by using a friendship network diagram, and questions which explain this diagram. I then obtained information about the identities in this specific context and their salient attributes (e.g. asking “What groups do you see in this area?”). Data is compiled in a corpus, the Language and Identity in a Dublin Suburb (LIDubS) corpus, which contains: (1) Interviews and free recordings, (2) Interview and observation reports (ethnographic information), (3) Transcription files, (4) Database of information about participants and the context, and (5) Analysis files. All tokens were distinguished (slit-t versus glottal variant) by ear auditorially, followed by verification using spectral analysis and acoustic analysis which can be used to distinguish realizations. The final data set included 816 word-final /t/ tokens. I now define some of the important independent variables:



Chapter 10.  Slit-t in Dublin English 163

a. The centre of gravity is a statistical measure, in Hertz, of the spectrum shape involving an amplitude-weighted mean of frequency components, or the average frequency of the spectrum overall (Jongman, Wayland and Wong 2000). b. The duration of the consonant is calculated from the start of the consonant (after the last burst of the preceding vowel, which can be judged by F2 climax of diphthong, or where formants disappear) to the end of the consonant: where final frication noise or glottal pulse appears in the spectrogram, or in other cases before the beginning of a new articulation. c. There are three categories for the position in the Intonational phrase variable: initial, medial and final segment of the intonation unit. d. The binary Pragmatic Class predictor involves coding each token whether it is a PM, or not. I analysed how the word-final /t/ variants are distributed across the pragmatic functions: a lexical function (not a PM), the information indicator function, and subjectively expressing a response to preceding discourse (Brinton 2008: 1). e. The coding of the Talk Type predictor featured three categories of talk prevalent in the interview data, as discussed above: Dressing Room Talk, Information Talk, and Social Talk. 3. Analysis of word-final /t/ This section presents analysis that suggests the use of a slit-t is indexed to an authoritative interactional identity. I first introduce the results of mixed effects modelling, which identifies factors that significantly predict variation in word-final /t/. I then outline some summary statistics for these predictors and highlight the salient patterns. Statistical analysis examined the relationship between variation in the realisation of word-final /t/, the binary (glottalised or slit-t realizations of word-final /t/) response, and three types of predictors: linguistic, social attributes, and interaction. From the total of 46 predictors, 11 predictors (in bold text in the list of predictors in Appendix 1) were selected by the least angle regression analysis (LARS) in the best fit model for a generalized linear mixed effects model (GLMER) analysis. I conducted general-to-specific mixed effects modelling, until an optimum model is reached in which all predictors are significant, or nearly significant. The results indicate that there are two significant predictors of variation in word-final /t/: the interaction/stance variable of Talk Type [p = 0.00179], and the linguistic variable pragmatic class [p = 0.032]. The linguistic variables duration (of consonant) [p = 0.05837], position in intonation phrase, [p = 0.0617], and centre of

164 Fergus O’Dwyer

gravity [p = 0.06401] were ranked as being nearly significant. While Talk Type and pragmatic class are both sociopragmatic in nature, it is suggested that the linguistic variables interact with pragmatic matters in this data. Statistical results serve as a heuristic to discover what predicts variation, and identify a unit or units of analysis for more in-depth analysis (Holmes 2014: 185). I now examine patterns relating to the above predictors in more depth. 3.1

Salient patterns of variation in word-final /t/

I first compare figures for slit-t and glottal realizations in order to identify salient differences, and to ascertain if the two variants behave differently. I then present details for the two nearly significant continuous predictors (duration, and centre of gravity of the consonant), before summarizing the corresponding binary predictor (Position in intonational phrase). I finally discuss the two statistically significant predictors – Talk Type, and the Pragmatic class (whether a PM or not) – before summarizing the main issues that emerged. The main findings are that slit-t tokens are often intonational phrase position-final and feature a lengthy frication, which seems to do sociopragmatic work like emphasize the speakers’ opinion. Importantly there is strong relationship between the slit-t variant and both Information Talk and PM tokens. Overall, this indicates the slit-t variant is used to indicate epistemic status: to indicate epistemic status, so as to establish, for example, a position of knowledge or authority. In regards to the figure for the centre of gravity (the average frequency of the spectrum overall), the central tendencies of glottal realizations are much lower than slit-t: this is to be expected as fricatives contain more energy at higher frequencies. Regarding duration, Table 1 shows that the slit-t tokens are typically twice as long as glottal variants, with a lot more variability in the slit-t tokens. Table 1.  Duration of all word-final /t/, glottal and slit /t/ tokens Duration (in seconds)

All word-final /t/

Glottal

Slit /t/

Median Mean Standard Deviation (SD)

0.04988    0.05993    0.04518151

0.04639    0.04927    0.02447923

0.097      0.1147     0.07680748

As many of the glottal variants are the result of a loss of oral articulation (an alveolar stop is the standard pronunciation in the case of the word-final /t/ of that, for example), this is to be expected. However, the duration of the slit-t tokens is remarkable, and it is possible that the extension of frication fulfils some sociopragmatic

Chapter 10.  Slit-t in Dublin English 165



Cumulative percentage

functions. The duration of frication for the PM token in Excerpt 2 below ‘lads who went like to Area 1 School right’ is 0.589 seconds. I argue that such a lengthy frication is done for non-phonetic reasons: to denote sociopragmatic weight. The analysis in the following section expand upon this argument, showing that use of the slit-t variant behaves differently according to sociopragmatic function. A related topic is the duration of the pause after the fricated word-final /t/. My analysis of a small selection of slit-t tokens (O’Dwyer 2019: 248) found that there is a median pause of 0.2885 seconds, a mean of 0.6394 seconds, with a standard deviation of 0.6077793 seconds. In the case of Excerpt 3 – “But (pause) I suppose the biggest things with our club” – and Excerpt 4 – “which isn’t great (pause) you know” examined below, there is a pause approaching two seconds, before finishing the sentence. Such a long pause in connected speech is unusual in the interviews and generally in Ireland: a lengthy pause following a word-final slit-t could possibly function as an implicit suggestion by the speaker that the interlocutor should pay attention, as there is “more-to-come” (Selting 2000: 512). This is one of the points covered by the discourse analytically-informed analysis in the next section. Another relevant measure includes position in the intonational phrase. While all the tokens are word-final, the majority (81.5%) of the tokens are found in medial positions of the intonation phrase, see Figure 1. 100

Final Medial Initial

80 60 40 20 0

All word-final /t/

Glottal

Slit /t/

Figure 1.  % phrase positions by types of /t/ variants

37.59% of slit-t tokens are position-final. This is in great contrast to much lower figures overall (8.63%), and for glottal realisations (13.35%). It is possible to infer that there is a relationship between intonational phrase position final slit-t tokens, and sociopragmatic work like emphasizing the point of discussion, for example. The remaining two predictors, which are both ranked to be statistically significant predictors of variation, also relate to epistemic matters. I analysed an equal amount of Dressing Room Talk, Information Talk and Social Talk tokens in the data set.

Cumulative percentage

166 Fergus O’Dwyer

100

Dressing Room Information Social

80 60 40 20 0

All /t/ tokens

Glottal

Slit /t/

Figure 2.  % /t/ variants by types of talk

Cumulative percentage

I hypothesized that as Dressing Room Talk is dominated by glottalised variants, this variant is indexed to a tough persona. The figures for all word-final /t/ tokens for the glottal realizations indicate Dressing Room Talk make up 36.31% of total glottal tokens, while Information Talk makes up 29.58% of total glottal word-final tokens. There is a remarkable contrast in the figure for slit-t tokens: only 18.79% of these tokens are coded as Dressing Room Talk, while Information Talk makes up 45.86% of slit-t tokens. These differences are statistically significant (χ² = 6.5518, df = 4, p-value = 0.0416). The relatively high percentage of slit-t tokens in Information Talk indicates an indexical relationship between slit-t and epistemic status. There is strong relationship between this variant and discourse where epistemic status is foregrounded. I analysed how the word-final /t/ variants are distributed across the following pragmatic functions: a lexical function (not a PM), the information indicator function, and subjectively expressing a response to preceding discourse (Brinton 2008: 1). Figure 3 presents the analysis of each function in PM tokens (i.e. all right, right, and but tokens). 100

Lexical Information Subjective

80 60 40 20 0

All PM tokens

Glottal

Slit /t/

Figure 3.  % of pragmatic functions by types of /t/ variants

There is a far greater percentage of pragmatic markers in the slit-t token figures, in comparison to the overall and glottal figures. A large number (57%) of slit-t PM tokens subjectively express an attitude to preceding discourse, with relatively few fulfilling a lexical function. This indicates that PM tokens are used to indicate epistemic status: in the case speakers use a PM – all right and but in particular – to establish they are in a position of knowledge or authority. This analysis provides



Chapter 10.  Slit-t in Dublin English 167

corroborating evidence that suggests a slit-t indexes epistemic status. The slit variant is more frequent in pragmatic marker tokens that express attitudes, while the glottal variants are commonly found in lexical functions. The evidence suggests that slit-t tokens in this corpus are used differently in comparison to glottal variants in terms of epistemic-related matters. The analysis so far suggests that there is an indexical link between a fricated word-final /t/ and utterances of sociopragmatic significance. The distribution of slit-t variants is heavily skewed toward Information Talk (discourse where epistemic status is foregrounded), with a relatively large relative number of not only pragmatic markers, but also position-final in the intonation phrase. Overall, this implies that the slit-t variant fulfils important sociopragmatic functions. The following section interprets representative examples of slit-t tokens, in order to clearly outline the social meanings and discourse functions of this variant. 4. The discourse functions of the word-final slit-t variant Four principal discourse functions of a slit-t are found in my data: to indicate speakers will do something sociopragmatic (e.g. subjectively express attitude), to introduce or make a point of sociopragmatic weight, to do sociopragmatic work (with a pause), and to emphasize a point (i.e. position themselves in authority or knowledge). While these functions are similar, they differ in terms when the sociopragmatic force is introduced temporally and sequentially, in relation to the slit-t realization. I use a discourse analytical-, ethnographically-informed interpretative framework to highlight how individuals position themselves socially, and interpret the authoritative interactional identities that are performed alongside use of slit-t realizations. For each excerpt examined below, I explain briefly the general background, then outline the relevant identities, how these identities are managed, and how the interaction is characterized as epistemic. In general, excerpts are selected in order to draw extensively from the most representative examples (Rosenwasser and Stephen 2011: 220). My overall aim is to interpret patterns of variation to create a better understanding of the relationship between linguistic variants and social meaning. I deal with each function in turn. 4.1

Slit-t indicates speaker will adopt a sociopragmatic position

This discourse function of slit-t precedes speakers doing sociopragmatic work (e.g. expressing an attitude). The [t̞ ] seems to indicate that the listener should pay attention to the attitude the speaker will go on to articulate. This is often in reply to a question from me in the interview and gives some more thinking time to the

168 Fergus O’Dwyer

speakers to formulate a considered response, but also indicates they have an attitude or position of knowledge to express. It is a type of a pre-sequence: the speaker wants to be given the opportunity for an extended multi-unit turn (Selting 2000: 484, 512; Sidnell 2010: 103). In the following examples, I provide a brief overview of the topic, then outline the sociopragmatic work, a social judgement for example, and how the slit-t token (underlined) relates to this social judgement. Excerpt 1.  I ask LiDub104 about the speaking styles of his friends 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

LiDub104: FOD: LiDub104: FOD: LiDub104:

FOD: LiDub104: FOD: LiDub104:

I just think there is a few different Dublin accents knocking around like, you know. Yeah, yeah so within this group how many accents would there be? I’d say like right let’s see. Well these are all country fellows. These two? These two. Well I’d say now you could put one, two two of these now by listening to them you’d think they are from the top of the Ballymun flats, you know, but they’re not like, you know, sure it doesn’t matter where you are from anyway. That’s [friend #1]? [Friend #1] and [friend #2] right. One, two, three of them you could think they were from Foxrock, Dublin 4 like, you know. Foxrock yeah. Like the kind of finely spoken. The other lads, who did I say now? This fellow here now he would be a real inner city Dub like the accent he has like.

When I first ask LiDub104 about the speaking styles of his friends (line 3–4) in Excerpt 1, his first reaction (“I’d say like right let’s see”, line 5 clearly gives him some more thinking time, but also indicates he can and will make social judgments on his friends, which he does in line 5 (“country fellows”) and line 9–10 (by listening to their accent, people think some of his friends are from the Ballymun flats). The second slit-t (line 14) is found where I try to continue the topic by further querying about friends. LiDub104 again indicates he will do some sociopragmatic work, ending in a social judgment (line 15–16: “you could think [they were from Foxrock] Dublin 4 like”; and line 18: “Like the kind of finely spoken”). In the social ideology of Dublin, these two locations represent the two types of opposing identities discussed above: Ballymun is an area typically featuring disadvantage, and relates to the proper identity, while Foxrock is one of the most affluent areas in Dublin and is indexed to the posh identity. LiDub104 situates himself in a position of knowledge about the social dynamics of the area and also someone who is in between proper and posh identities. Similar identity work is found in other excerpts also (see O’Dwyer 2019). In these examples, the word-final slit-t seems to suggest there

Chapter 10.  Slit-t in Dublin English 169



is something worth listening to in what follows: the speaker is projecting there is more to follow. 4.2

Slit-t introduces a point of sociopragmatic weight

This discourse function involves the speakers introducing a point which clearly expresses a new attitude or position, by using a word-final slit-t. Speakers then go on to further elaborate upon the reason for this attitude. It is a type of second pair part, with expansion, that “promotes the accomplishment of the activity underway” (Schegloff 2007: 124–125). The word-final slit-t seems to introduce the sociopragmatic weight of their argument, in order to get the attention “… of the listener and to attain a position of authority etc.”. In the explanations below, I try to outline how an argument is introduced – with the slit-t denoting significance – before explaining how the speaker further elaborates on this point. Excerpt 2.  I ask LiDub104 about the ways of talking in Club Fingal 1 FOD: 2 3 4 LiDub104: 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

What about when you see you’d have different lads down the club would you see different ways of talking from different lads down the club? Oh I would yeah, yeah. For instance now, this would be fact. Like this is my personal opinon anyway. Any of the fellows let’s say who went like to [Area 1] National School right would speak in a more, in a posher accent I think as opposed to like the lads like any of us who went to school in [Area 2]. Now don’t ask me, we have more prominent Dublin accents like, you know, whereas I always thought that like a lot of the lads were more kind of, they spoke with a small bit of a posher tone kind of thing like. A few more niceties about them or something like. Now I could start mentioning a few names! It’s not a slag. It was just the way I always kind of thought that like, you know, and I remember playing soccer years ago playing against [Area 3 soccer team name], you know, and like just the language they use was something completely diffeent to what we used like

One of the main claims in this interaction is introduced somewhat hedgingly (“my personal opinion anyway” and “let’s say”) in line 5–7: those who went to Area 1 National School (i.e. school for 5–12 year old children) speak with a posher accent. This claim is followed by the PM right (line 7), which seems to indicate that there is sociopragmatic weight attached to the point which LiDub104 introduces. By looking back from the end of this interaction (line 20), it is clear that LiDub104 feels he has a lot to say about this topic, and the right token (line 7) announces this. LiDub104 then elaborates in lines 7–13 how those from the Area 1 speak and act posher, while

170 Fergus O’Dwyer

the language of people from Area 3 is completely different (i.e. not posh, line 20). Here the speaker again places himself in a position of knowledge and experience to make such judgements, and also as an Area 2 person “in between” those identified with the Area 1 school (posh) and Area 3 soccer team (proper) identities. 4.3

Slit-t does sociopragmatic work, with a pause

I interpret that speakers in this data often do sociopragmatic work, such as subjectively expressing an attitude etc., by using a slit-t variant followed by a pause. This word-final slit-t is in between one clause that presents their attitude, and a following main clause that further expresses their attitude (or does similar sociopragmatic work). I view this as a type of an insertion sequence, which delays the second pair part to deal with issues that need to be resolved before the second pair part can be produced (Sidnell 2010: 103). An insert expansion may occur, for example, as an attention-holding device, for the projection of “more-to-come” (Selting 2000: 512). In this case, the speaker is in the middle of conveying a point, then expresses sociopragmatic weight by realizing a slit-t, before going on to further express their attitude (possibly in a different way), and further elaborate on their opinion. The pause seems to function as an indicator that although the speaker has expressed an attitude, there is a more important point that remains, that they have not yet fully communicated. Excerpt 3.  LiDub106 is discussing social life in Club Fingal 1 FOD: 2 3 4 LiDub106: 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

So, how often would you socialise with the lads on the team and what would you do, what would be the social life? Well I suppose it is very difficult to socialise, I would classify sharing and having the craic in the dressing room after training as being social. We spend a half an hour or 40 minutes talking in the dressing room and slgging the lads and talking about the weekend and what happened or something in the news or politics or anything. Then I suppose bigger social occasions would be things that obviously if you won something- you won an important match obviously we would go to the bar afterwards and you would have a few drings and you would chat there. But (2.0) I suppose the biggest things with our club would be at the start of every season, each team, or each group of players would have a, there would be an organised trip away Yeah. And for the last seven or eight years since I am involved with the teams we would go to different parts of the country and 30 lads would head off for the weekend.

Chapter 10.  Slit-t in Dublin English 171



LiDub106 develops in an extended dialogue, from line 4–14 of Excerpt 3, his views about socializing in Club Fingal. He expresses the sociopragmatic weight through the PM but (line 14), and then further conveys his attitude by elaborating on the related point of pre-season trips away in line 14–21. The PM-final slit-t is at the centre of LiDub106s’ lengthy explanation about the social life of the club. He first defines the social activity of chatting, and then – pausing after but – LiDub106 goes onto explain to me, the unknowing, about the trips away, and how they are central to social bonds. He positions himself as one who has knowledge and experience of the inner workings of a club, and what socialising in the club really involves. In this case, the slit-t is both a lexicalised marker of “more-to-come”, while it also does the work of asserting epistemic status. In other similar examples (O’Dwyer 2019: 257), speakers emphasize their position as someone in a position of authority, to explain the recent history of the club, how involved he is in these practices, and how rewarding it is, for example. 4.4 Slit-t to emphasise a point This discourse function involves the speaker emphasise a point by realising a slit-t at the end of a sequence where they have expressed their attitude. This is a type of post-sequence, which expands on the base sequence after the second pair part (Sidnell 2010: 103). Post-expansions can be “sequence-closing thirds”, with which the speaker intends to close a sequence. The utterance- and sequence-final slit-t seems to serve the purpose of emphasising their point of view, and authoritative position. Excerpt 4.  LiDub107 is discussing his friends 1 2 3 4 5 6 7 8

FOD: LiDub107:

Where is he, over in Oz or? In Australia, he was kind of between Sydney and Melbourne and then he spent a couple of months doing the East Coast. And then we had two other guys who werein trades and couldn’t get work in Dublin. In fact at one stage they were working in belmullet in Mayo just to keep themselves in a job which isn’t great (2.0) you know.

After an extended discussion of a bad situation regarding emigration, LiDub107 concludes his position on this matter with an assessment (“which isn’t great” in line 7), there is a pause of 2 seconds before the pragmatic marker “you know” which serves the purpose of asking for confirmation or agreement. The utterance-final slit-t possibly emphasises his position of empathy regarding his friends’ predicament, but also positions the speaker as someone who is lucky to be in steady employment, occupying a position of knowledge about the unemployment problem in Ireland (in comparison to me).

172 Fergus O’Dwyer

5. Conclusion: Slit-t in Dublin English The evidence outlined above strongly suggests that the LIDubS Corpus contains many instances of speakers using a slit-t variant when adopting authoritative interactional identities. The previous section described four discourse functions of slit-t, all of which are linked to speakers performing sociopragmatic work such as subjectively expressing attitudes. Future work in other contexts in Dublin and beyond can examine if similar patterns are used elsewhere, with any variation. In general though, my results provide locally-informed information about the social dynamics of language variation. An understanding of how phonetic variation is produced and performed should involve the integration of ethnographic approaches, to investigate the indexical meaning of particular phonetic variants (Hay and Drager 2007: 90). The analytical steps adopted here present an example of how ethnographically informed, qualitatively-skewed mixed methods can elucidate the social meaning of linguistic variants. While a focus on the discourse functions of sociophonetic variants is not unique (see Podesva 2007 for example), this paper presents a thorough, systematic and easily-implemented method for the investigation of social meanings: first statistical results highlight the patterns of variation that deserve more attention and interpretations. The detailed, qualitative analysis then leads to a better understanding of the relationship between linguistic variants and social side of interactions.

References Amador-Moreno, Carolina P., Kevin McCafferty and Elaine Vaughan. 2015. Introduction. In Carolina P. Amador-Moreno, Kevin McCafferty and Elaine Vaughan (eds.), Pragmatic markers in Irish English, 1–16. Amsterdam: John Benjamins Publishing. Brinton, Laurel J. 2008. The comment clause in English. Cambridge: Cambridge University Press. Carter, Ronald and Michael McCarthy. 2006. Cambridge Grammar of English: A comprehensive guide. Cambridge: Cambridge University Press. Clancy, Brian and Elaine Vaughan. 2012. It’s lunacy now: A corpus-based pragmatic analysis of the use of “now” in Contemporary Irish English. In Bettina Migge and Máire Ní Chiosáin (eds.), New perspectives on Irish English (G44: Varieties of English around the world), 225– 246. Amsterdam: John Benjamins Publishing. Eckert, Penelope. 2012. Three waves of variation study: The emergence of meaning in the study of sociolinguistic variation. Annual Review of Anthropology 41 (1). 87–100. https://doi.org/10.1146/annurev-anthro-092611-145828

Hay, Jennifer and Katie Drager. 2007. Sociophonetics. Annual Review of Anthropology 36 (1). 89–103.  https://doi.org/10.1146/annurev.anthro.34.081804.120633 Hickey, Raymond. 2005. Dublin English: Evolution and change. Vol. G35. Varieties of English around the world. Amsterdam: John Benjamins Publishing.



Chapter 10.  Slit-t in Dublin English 173

Holmes, Janet. 2014. Doing discourse analysis in sociolinguistics. In Janet Holmes and Kirk Hazen (eds.), Research methods in sociolinguistics: A practical guide, 177–93. Oxford: WileyBlackwell. Jongman, Allard, Ratree Wayland and Serena Wong. 2000. Acoustic characteristics of English fricatives. Journal of the Acoustical Society of America 108 (1). 252–63. Kallen, Jeffrey L. 2013. Irish English Volume 2: The Republic of Ireland. Berlin-Boston: de Gruyter. Kirk, John M and Jeffrey L. Kallen. 2011. The cultural context of ICE-Ireland. In Raymond Hickey (ed.), Researching the languages of Ireland, 269–90. Uppsala: Uppsala University Press. Lonergan, John. 2013. An acoustic and perceptual study of Dublin English phonology. Unpub­ lished PhD Thesis. University College Dublin. O’Dwyer, Fergus. 2019. Language and interactional identity in a Dublin suburb. Unpublished PhD Thesis. University College Dublin. Omoniyi, Tope. 2006. Hierarchy of identities. In Tope Omoniyi and Goodith White (eds.), The sociolinguistics of identity, 11–33. London: Continuum. Podesva, Robert J. 2007. Phonation type as a stylistic variable: The use of falsetto in constructing a persona. Journal of Sociolinguistics 11 (4). 478–504. Rosenwasser, David and Jill Stephen. 2011. Writing analytically. 6th edition. Boston, MA: Wads­ worth Publishing. Schegloff, Emanuel A. 2007. Sequence organization in interaction. Volume 1: A primer in conversation analysis. Cambridge: Cambridge University Press. Selting, Margret. 2000. The construction of units in conversational talk. Language in Society 29 (4). 477–517. Share, Bernard. 2006. Dublinese: Know what I mean? Dublin: Collins Press. Sidnell, Jack. 2010. Conversation analysis: An introduction. Oxford: Wiley-Blackwell.

Appendix Linguistic considerations, social attributes and interactional/stance variables, which are used as predictors in statistical analysis. Linguistic considerations: Duration (ms), Pitch/f0 max (Hz), f0 range (max – min), f0 Mean, Intensity max (db), relative intensity (mean vowel – mean /t/); Centre of gravity, standard deviation, skewness, kurtosis and central moment; Preceding vowel, Following Manner, Following Place, Following Voiced (Y/N), Stressed syllable (Y/N), Position in intonation phrase, Number of syllables in word and Intonational phrase, Word class, Pragmatic Class (Pragmatic Marker Y/N). Social attributes: “Attitudes” (toward: the club, criminality, individuals not typically associated with the ‘pillar’ market, rural life, suburban life); “Background” (Birth heritage of parents, Education of father, Education of mother, Involvement of parents in the community, Occupation of father, Occupation of mother, Secondary school attended, Where they lived in formative years); “Lifestyle choices” (Fashion, Friendship Networks, Involvement in the club, Occupation, Sporting Choices, Where they socialise, Recreational choices). Interaction variables: Type of talk (Dressing Room Talk, Information Talk, Social Talk); Face-threat dimension (Levon and Holmes-Elliot 2014: Face threat to others, Reacting to a face threat, Face Strategy etc.); Topic (Family/friends, Club-County, social etc.), Delivery (relaxed – intense/animated); Direction (toward outside [e.g. other clubs/groups] – inside, others etc.); Orientation (Derogatory – Complimentary); Organisation (temporal place in interaction).

Chapter 11

Panel and trend studies Evidence from Brazilian Portuguese Maria da Conceição de Paiva and Maria Eugênia Lammoglia Duarte Federal University of Rio de Janeiro/CNPq

In this paper, we provide further evidence to the discussion of the intersection of apparent time and real time for the understanding of language change. We analyse three variable processes in the spoken variety of Rio de Janeiro: the anterior glide deletion in the diphthong [ej], the replacement of the preposition A by PARA (‘to’) in dative complements and the loss of null referential pronominal subjects. By combining cross-sectional and longitudinal evidence, we show that the observed tendency in apparent time is confirmed for the replacement of A by PARA and the implementation of overt referential pronouns. We are facing continuing and irreversible communal changes. In the case of [j] deletion, the predictions obtained in apparent time are only partially confirmed. The results from the longitudinal analyses suggest a reversal of this process in the context of palatal fricatives both in the community and in most individuals. We claim that this reversal is a generalisation of a stronger tendency to glide insertion in similar phonetic contexts. Keywords: glide deletion, dative PPs, null/overt subjects, apparent time, short term real time

1. Introduction The apparent-time construct – i.e. the use of synchronic evidence to ascertain the existence of changes in progress – has meant an important contribution of variationist studies and has proved to be a good hypothesis to track language change. The use of a linguistic variant by the younger age-group suggests an ongoing change, or more precisely, a generational change, if we accept the hypothesis of the stability of the individual linguistic system after adolescence. The problem, as pointed out by various authors (Labov 1994; Chambers 2002; Bailey 2002; G. Sankoff 2006) is that regular synchronic distributions of a linguistic variant across different age cohorts https://doi.org/10.1075/silv.22.11dep © 2019 John Benjamins Publishing Company

176 Maria da Conceição de Paiva and Maria Eugênia Lammoglia Duarte

can also indicate an age-grading pattern, i.e. changes in the speech of older speakers which are repeated in each generation, while the community remains stable. Evidence from apparent time studies can be checked by real time analysis of the community (trend study) and the individuals (panel study). The combination of these techniques shows that the interrelation between change in the community and change in the individual is complex since the existence of a regular effect by the speaker’s age could also indicate a lifespan change. In other words, some speakers adopt new linguistic variants, following the direction of a change in progress in the rest of the speech community. According to Sankoff (2006) and Sankoff and Blondeau (2008), continuing change and age-grading can be interrelated and in this case two situations can be identified: if there is no change, the more prestigious variant is associated with the intermediate group of adult speakers (showing a curvilinear pattern), presumably in response to linguistic market pressures. In the case of ongoing change, the new variant can, to some extent, be adopted by older speakers. In opposition to age-grading, lifespan change is not cyclic. Table 1 sums up the possible interpretations of regular age distributions proposed by Sankoff and Blondeau (2008, 563), where a supplementary pattern reflecting lifespan change at the community level has been included. Table 1.  Age distributions. Source: adapted from Sankoff and Blondeau (2008) Synchronic pattern

Interpretation

Individual

Community

Flat Regular slope with age Regular slope with age Regular slope with age (apparent-time interpretation) Flat

1. Stability 2a. Age-grading 2b. Lifespan change 3. Generational change 4. Communal change

stability change change stability

stability stability change change

change

change

A clear example of life-span change is the spread of dorsal [R] in Montréal French. (G. Sankoff 2006: 1011). The trend study shows a great increase of the dorsal variant and a detailed panel study allows the identification of a group of re-interviewed speakers that changes in the same direction, increasing their use of the new variant or adopting it. This paper returns to this subject, discussing some cross-sectional and longitudinal evidence for three widespread variable phenomena in the spoken Portuguese of Rio de Janeiro. Our objective is twofold: the first is to provide additional evidence that apparent time predictions can overestimate the spread of linguistic changes in the community. The second is to discuss to what extent change in the linguistic behaviour of the individuals is related to the type of change (change from below or change from above) and its degree of propagation in the speech community.

Chapter 11.  Panel and trend studies 177



2. The phenomena A large number of studies, based on synchronic evidence of Rio de Janeiro Portuguese, has allowed the identification of patterns of variation compatible with ongoing changes for several variable processes in this city (see Oliveira e Silva and Scherre 1996). In this paper we focus on three of these phenomena which are near completion and can be characterised as changes from below. They are:

(1) The deletion of the anterior glide [j] in the diphthong [ej] as in beijo ‘kiss’: [ˡbejʒʊ] ~ [ˡbeʒʊ]

(2) The replacement of the preposition a ‚‘to’ by the preposition para ‘to’ heading dative PPs. Vou dar um presente a ele. → Vou dar um presente para ele. ‘I’ll give a  present to him’.

A third variant is possible: the suppression of the preposition (Ele deu um presente o Papa ‘he gave a present to the Pope’ or Ele deu o papa um presente (‘He gave the Pope a present’). This variant is subject to diatopic variation, AND very unfrequent in Rio de Janeiro speech and will not be discussed in this paper. (3) The loss of null referential subjects. Minha esposa trabalha na Embratel. Ela fez segundo grau técnico em contabilidade. Depois ela fez faculdade. Hoje ela é técnico em contabilidade da Embratel. Ela ganha bem, mas eu acho que ela devia ganhar mais porque ela merece. ‘My wife works at Embratel. She took a course in accounting at a secondary technical school. Then she went to college. Today she is an accounting technician at Embratel. She earns well, but I think she should earn more because she deserves it.’

The findings discussed in this paper come from a combination of trend and panel studies carried out in the city of Rio de Janeiro. The trend study is based on two samples (Sample 1980 and Sample 2000) collected in the urban area of the city with an interval of about 20 years. Sample 1980 was recorded between 1980–1983, with 64 native speakers (see Oliveira e Silva and Scherre 1996) and Sample 2000, recorded in 1999–2001, with 32 speakers (see Paiva and Duarte 2003).1 The two samples were stratified according to: four age groups (7–14, 15–25, 26–49 and over 50 years); years of school attendance: (1 to 4; 5 to 8; 9 to 11 years), and gender. To analyse the relation between change in the individual and the community, 16 individuals who had participated in the first survey were re-interviewed for the panel study.

1. These two samples are available in the following address: http://www.letras.ufrj.br/peul/.

178 Maria da Conceição de Paiva and Maria Eugênia Lammoglia Duarte

3. Deletion of the glide [j] – evidence from apparent and real time The variable phonetic realisation of the anterior glide [j] in the diphthong [ej] can be considered as a change from below, since it is inserted in a general-drift that takes its root from Latin. This process is constrained by the position of the glide, which is categorically retained in final position, as in [xej] ‘king’, [faˈlej] ‘I spoke’. In medial position, the deletion of the glide [j] is blocked before vowels ([ˈmejɐ] ‘socks’), dental stops ([ˈpejtʊ] ‘breast’), alveolar fricatives ([ˈbejsʊ] ‘kiss’) and laterals ([lejˈlãw] ‘auction’). [j]-deletion is particularly productive before palatal fricative consonants as in [ˈbejʒʊ] ~ [ˈbeʒʊ] ‘kiss’, [ˈpejʃi] ~ [ˈpeʃi] ‘fish’ or a flap as in [­maˈdejɾɐ] ~ [maˈdeɾɐ] ‘wood’. In other phonetic environments, such as before the velar voiced stop [g] ([mãˈtejgɐ] ~ [mãˈtegɐ] ‘butter’), the nasal bilabial ([ˈkejmɐ] ‘burn’) or the nasal alveolar [n] ([ˈtɾejnʊ] ~ [ˈtɾenʊ] ‘training’ ‘rehearsal’), the process is very rare and usually restricted to a single lexical item (Paiva 1996: 2003). Taking into account only the variable contexts, i.e. before the flap, palatal fricatives, the velar voiced stop and the bilabial nasal, Paiva (2003) concluded that the glide deletion in the diphthong [ej] had undergone an almost complete change at least since the 1980s, reaching 91% and 89%, respectively in Sample 1980 and Sample 2000, as shown in Table 2.2 Table 2.  Frequency of [j] deletion in 1980 and 2000 Sample 1980

Sample 2000

%

N

(T)

%

N

(T)

91

731

(795)

89

711

(797)

However, a logistic regression analysis, conducted with GoldVarb 2001 (Robinson et al. 2001), shows that, even if flap and palatal fricatives remain the most favorable contexts for glide deletion in the diphthong [ej], there are indications of a different direction in the effect of these two phonetic environments, as shown in Table 3. As shown in Table 3, even if there is a difference between the means of [j] deletion before the flap ([ˈfeɾɐ] ‘market’), and palatal fricative ([ˈkeʃʊ] ’chin’), there is no significant difference in the FWs associated with these two contexts in Sample 1980. In both environments this process can be interpreted as a nearly completed change with FWs of .89 and .82 respectively. In the second synchrony, however, the nature of a concluding change of [j] deletion in the flap environment (mean = 97%, FW .92) is confirmed although surprisingly this process decreases in the context of palatal fricatives (mean = 66%, FW .38). In Sample 2000, we can see that the latter 2. In the tables, T is the total number of tokens and N refers to the number of applications.

Chapter 11.  Panel and trend studies 179



Table 3.  Glide deletion according to the phonetic environment in sample 1980 and sample 2000* Structural context Flap Palatal Fricatives Velar stop Nasal bilabial

Sample 1980

Sample 2000

F.W.

%

N

(T)

F.W.

%

N

(T)

0.89 0.82 0.01 0.08

98 76 30 71

576 147   3   5

(585) (193)  (10)   (7)

0.92 0.38 0.31 0.34

97 66 67 70

578 124   2   7

(597) (187)   (3)  (10)

* The results in Table 3 were obtained from separate runs for sample 1980 and sample 2000.

phonetic environment becomes more significant as shown by the FW (.92 vs .38) and the higher range for this synchrony (.54). Consequently, we have evidence for a possible reversion of the process in the palatal fricative environment in the lapse of 20 years. A more detailed analysis of [j] deletion before palatal fricatives in both periods according to the speaker’s age reveals important differences between Sample 1980 and Sample 2000, as shown in Figure 1:3 100

1980 2000

80

(%)

60 40 20 0

7 to 14

15 to 25

26 to 49

> 50

Figure 1.  Glide deletion before palatal fricatives by age groups in sample 1980 and sample 2000

Figure 1 shows a flat pattern for age in the 1980s, which is compatible with an almost completed change. The glide deletion before palatal fricatives is uniformly spread among all the age groups, with near categorical rates (above 90%). In the 2000s, on the contrary, we obtain a curvilinear pattern: the highest rates of glide deletion occur in the first and fourth age groups and the low rates in the intermediate ones. 3. The first age group comprises 4 speakers, the second and the third one, 10 speakers and the older group (50 years old) comprises 8 individuals.

180 Maria da Conceição de Paiva and Maria Eugênia Lammoglia Duarte

The panel study reinforces the evidence of a reversal in the direction of glide deletion before palatal fricatives, as shown in Table 4. Taking into account that the education level is an important factor in the explanation of variation in BP, the 16 re-interviewed speakers of the Sample 1980 were gathered in two subsets: speakers who had and those who had not furthered their formal educational level by the time of the second interview. Table 4.  Means of [j] deletion before palatal fricatives in the individuals in sample 1980 and sample 2000 Speakers who furthered their formal education Speaker Eri59 Adr57 Adr63 Fat23 San39 Leo38

Gender woman woman woman woman woman Man

1980 Sample

 

2000 Sample

 

Age

%

N (T)

Age

%

N (T)

 9 10 12 15 15 18

100 100   96 100   50   63

4 (4) 3 (3) 22 (23) 7 (7) 1 (2) 10 (16)

25 26 28 33 33 36

  71   71   67   67   35   21

10 (14) 12 (17) 2 (3) 8 (12) 6 (17) 1 (5)

 80  80  80  88   31  75 100   25  50   71

16 (23) 4 (5) 20 (25) 14 (16)  4 (13)  9 (12) 9 (9) 2 (8) 3 (6) 5 (7)

Speakers who did not further their formal education Jup06 Lei04 Dav42 Jvas26 Eve43 Mgl48 Jan03 Nad36 Jos35 Ago33

woman woman man man woman woman man woman woman man

18 25 31 32 42 52 56 57 59 60

 80  78  62  38   67  50  90   88  50 100

40 (50) 7 (9) 8 (13) 5 (13) 8 (12) 6 (12) 38 (42) 7 (8) 3 (6) 13 (13)

35 43 48 48 59 70 74 74 75 77

Three situations are depicted in Table 4: (a) nine speakers (percentages underlined) decrease the rate of glide deletion before palatal fricatives; (b) four speakers (percentages in italics) increase their rates of suppression of the glide and (c) 3 (percentages in normal font) remain stable. Thus, the great majority of speakers in this group follows the direction suggested by the trend study. The different trajectories shown in Table 3 make it clear that the unexpected pattern attested for Sample 2000 is dependent on social factors. Six out of nine speakers that follow the tendency of the community are included in the subset of those who increased their level of schooling. Some of these individuals are graduates and were, therefore, more exposed to reading and writing. Five of these six



Chapter 11.  Panel and trend studies 181

speakers are women who show a greater decrease of glide deletion, around 30 percent points in the lapse of twenty years. In the subset of speakers who have not changed their level of education, only two women (Eve43, Nad36) and one man (Ago33) decrease their rate of glide deletion. All of them are retired and pertain to the middle or low-middle class. These facts suggest that the reversal of the glide deletion before palatal fricatives cannot be convincingly explained as an exclusive result of schooling. Furthermore, we must point out that 7 out the 9 speakers who decreased their rate of [j] deletion are women. The unexpected reduction of [j] deletion in palatal fricative environment in the third age group is conspicuous: how can we interpret it? First, we must take into account another phonological process concerning glide [j] in Rio de Janeiro variety. As it has been pointed out in some studies (Bisol 1989, 1994; Leite, Callou and Moraes 2003), the contemporary variety of Rio de Janeiro Portuguese shows a strong tendency to insert [j] in the context of palatal fricatives in stressed or non stresssed syllables, such as pasta ~ paysta ‘toothpaste’, festa ~ feysta ‘party, festança ~ feystança ‘big party’.4 In fact, it seems that this diphtongation process also spreads to the context of alveolar fricatives, as nascimento ~nayscimento ‘birth’.5 As shown by Leite, Callou and Moraes (2003), this diphtongation tendency in the speech of Rio de Janeiro is correlated to the variable gender. In a panel study comparing the speech of graduate speakers in the 1970s and 1990s, the authors show that in the lapse of twenty years women prese nted an age pattern compatible with generational change: diphthongisation before palatal fricatives was favoured by young women. We must also take into account that even if [j] insertion in these contexts is clearly not susceptible to social evaluation, this pronunciation can be considered as a phonological feature of carioca identity. As Rio de Janeiro is an important social and cultural centre, it can be assumed that variable insertion of the glide [j] has acquired covert prestige. If this is the case, we must explain how an unconscious linguistic variant has acquired social prestige. At first glance, we can hypothesise that the reversal observed in apparent time for Sample 2000 is embedded in a more general process of diphthongisation before palatal fricatives. If this hypothesis is true, the age distributions observed in apparent time for Sample 2000 can be initially interpreted as an age-grading pattern. This interpretation seems to be reinforced by the panel results. In the second synchrony, 4. We must point out that in the Rio de Janeiro variety, /s/ is realize as a voiced [ʒ] or voiceless [ʃ] fricative, in syllabic coda, according to the following segment. These are the realisations of the palatal fricatives following the glide in beijo and queixo, respectively. 5. It is important to make it clear that, even though European Portuguese also realises /s/ in syllabic coda as a palatal fricative, there is neither glide deletion nor insersion, which allows one to dismiss any possible influence of the European variety in the process reported here.

182 Maria da Conceição de Paiva and Maria Eugênia Lammoglia Duarte

the great majority of the speakers who decrease their rates of [j] deletion are women included in the third age group (between 25 and 36 years old), a suggestion that their adjustment in the direction of a linguistic form that acquired a covert prestige in Rio de Janeiro is a result of social pressures at the linguistic market. However, the reduction of glide deletion before palatal fricatives by three speakers who are over 59 years of age shows that the innovation is also adopted by older speakers, suggesting that we face a change in progress, possibly driven by other phonological processes and highly motivated by the gender variable. According to evidence drawn by Sankoff and Blondeau (2007), change during the lifespan can be continuous – centralised pattern – or more abrupt – polarised pattern – (see also Nevalainen et al. 2011; Baxter and Croft 2016). It seems that except for Ago 33, a male speaker, the individual change attested in the two female speakers could be considered as a more polarised change. While the average of [j] deletion in the group of the nine speakers who continued their studies has been reduced to 29.5%, the average of the three speakers who did not further their level of schooling is 42.6%. If we assume the position claimed by Baxter and Croft (2016), the sudden change observed in these speakers reflects their degree of accommodation: it is possible that these speakers have accommodated less readily to a tendency to reinforce a covert prestige feature of the community. 4. The replacement of the preposition A by PARA in dative complements – evidence from apparent and real time The replacement of preposition A by PARA in dative complements attested by Berlinck (1996), Gomes (1999, 2003), Freire (2000, 2005) among others, is also a widely spread phenomenon in BP. Just like the change in the subject expression, it seems to be embedded in more general changes triggered by a drastic restructuring in the Brazilian pronominal system. As expected in widely spread changes, the incrementation of preposition PARA is not strongly constrained by structural or social factors. It seems that only the semantic relation between the verb and the PP complement still resists the change: preposition PARA can appear with every type of verb, but it is outnumbered by preposition A with light verbs, particularly dar – ‘to give’ (dar continuidade a ‘to give contituity to…/dar valor a …– ‘to give value to…’), either with non-animate or animate referents. As we can see in the examples, the use of the preposition A is favoured in the order V DO IO. The trend study shows drastic reduction of preposition A in the second point of time (from 38% in the 1980s to 11% in the 2000s) and high incrementation of the preposition PARA (from 49% to 85%). The overall for the null preposition has been low since the 1980s (13%) and decreases to only 4% in the 2000s.

Chapter 11.  Panel and trend studies 183



100

A 1980 A 2000 Para 1980 Para 2000

80 (%)

60 40 20 0

7 to 14

15 to 25

26 to 49

> 50

Figure 2.  Frequency of A and PARA in dative PP by age in sample 1980 and sample 2000

The distribution of the two most frequent variants – A and PARA – according to the age groups, confirms the strong increasing of preposition PARA in dative PP in the speech of Rio de Janeiro, as displayed in Figure 3. 100

(%)

60 40

38

49

20 0

A Para Null prep.

85

80

13 1980

11

4 2000

Figure 3.  Frequency of A and PARA and null preposition in 1980 and 2000

In the 1980’s, the age pattern observed for the prepositions PARA and A in dative complements shows a clear opposition between the extreme age groups: preposition PARA is more frequent (64%) in the younger age group (7–14 years old) and decreases (39%) among the older speakers (> 50 years old), in contrast to a converse distribution of A preposition. In the intermediate groups (15–26 and 26–49 years old) there is more variation between PARA and A prepositions.6 The most interesting movement can be observed in the results displayed for the second point of time (Sample 2000): all age cohorts increase the use of PARA in dative complements, suggesting extensive incrementation of the change. There is no difference between the first and second age cohorts, with nearly categorical values. Even if there is a small difference between the means for the two younger groups and the older ones, age-group is discarded in the 2000´s. Therefore, we can conclude that the replacement of the preposition A by PARA is a communal change.

6. Unfortunately, the significance and FW values could not be recovered.

184 Maria da Conceição de Paiva and Maria Eugênia Lammoglia Duarte

Even speakers over 50 years old, who still present the higher rate of the preposition A, adopt the innovative form. A controlled panel study of the incrementation of the preposition PARA is more difficult, since the number of dative PP by speakers is very scarce. However, the comparison of the two interviews shows that the occurrence of preposition A is very rare in the two moments. In all the 16 re-contacted individuals, the head of dative PP is more often filled by PARA and this tendency increases regularly in the 2000s. This regularity reinforces the interpretation of a nearly completed communal change. 5. Overt (vs null) subjects in apparent and real time We now turn to the variation between null and overt subjects in the Rio de Janeiro variety. According to Duarte (2000, 2003), the tendency to overt pronominal subjects is an instance of change in progress during a longer period of time. This process is related to the reduction of the inflectional verbal paradigm, as a consequence of other two changes. The first concerns important changes in the set of personal pronouns: the full grammaticalisation of the former address form Vossa Mercê (‘your grace’), as você (‘you’), which enters in competition with pronoun tu (‘you’), and of the NP a gente (literally (‘the people’), which entered the system in alternation with nós (‘we’) and has already overtaken it (cf. Lopes 2002; 2003). Since both new pronouns combine with the unmarked 3rd person singular verb form, the BP verbal inflectional paradigm was significantly reduced. The second is related to a more general phonological process, the apocope of the phonetic segments [s] and [m], as in tu fala(s) ‘you speak’, eles fala(m) ‘they speak’, which aggravated the impoverishment of the inflectional paradigm. The variation null/overt pronominal subject in BP is constrained by structural factors as animacy, and clause type (embedded vs non-embedded) associated to the position/function of the antecedent. Animacy explains the faster implementation of overt subjects in the 1st and 2nd person forms as in (4) and (5): (4) Mesmo que eu não fizesse o pré-vestibular, eu acho que eu even if I not take-SUB the preparatory course I think that I passaria por causa da base que eu tinha. would-pass because of-the basis that I had. ‘Even if I had not taken the preparatory course, I think I would have passed because of the basis I had’ (5) Você me disse que você está morando em Copacabana. you cl–1ps told that you are living in Copacabana ‘You told me that you are living in Copacabana’

Chapter 11.  Panel and trend studies 185



The 3rd person shows more variation between overt and null subjects, even though rates of overt human subjects are ahead of non-animates. We must clarify that since BP does not have a neuter pronoun for non-animate referents, the pronouns he/she can refer to animate as well to non-animate referents, as in Example (6): (6) mas elei sentiu [que elei / Øi era o único novo ali, recém– casado]. ‘but hei felt [that hei/ Øi was the only young there, ’ newly– married] ‘But he felt he was the only young guy there, newly married’ (7) [Esse filme]i emocionou muita gente quando elei / Øi ficou pronto [That film]I touched many people when hei / Øi was ready ‘That film touched many people when it was shown’

As expected in relation to widely advanced processes of change (Labov 1963, 1972, 1994), there is no difference between the average for overt referential subjects, including, naturally, 1st, 2nd and 3rd person forms (animate and non-animate referents), in the two points of time, as we can see in Table 5. Table 5.  Frequency of overt (vs null) referential subjects in sample 1980 and sample 2000 Person 1st 2nd 3rd Total

Sample 1980

Sample 2000

%

N

(T)

Person

%

N

(T)

77 88 79 79

1722  157 1329 3208

(2212)  (178)  (1664) (4054)

1st 2nd 3rd Total

79 78 80 79

1910  130 1310 3350

(2409)  (166) (1635) (4210)

Two interpretations can be drawn from such results. The first is that there is no change in the lapse of 19 years. The second interpretation is that the actuation of the change towards overt subjects is close to completion, as the new form is adopted by the whole community, and the process slows down as modelled in the S curve. (See Labov 1963, 1972, 1994, Tagliamonte 2011). The second interpretation seems more plausible, if we consider the age correlations obtained for each synchrony, shown in Table 6. First, we must point out that the high percentages of overt subjects in all age groups, outnumber three quarters of the data in both samples. Moreover, the differences among age groups are statistically significant in the 1980s as well as in the 2000s, although selected in different levels: in 1980, age is selected in the fourth level, after three structural factors (syntactic accessiblity of the antecedent, embedded or non-embeded clause and presence vs absence of light elements between subject and verb). For the second synchrony age variable is selected in the ninth level.

186 Maria da Conceição de Paiva and Maria Eugênia Lammoglia Duarte

Table 6.  Frequency of overt subjects according to age-group in sample 1980 and sample 2000 Sample 1980 Age

F.W.

07–14 .51 15–25 .57 .49 26–49 >50 .41 Input 0.82 Range: .16

%

N

77 1722 88  157 79 1329 79 3208 Signif = .0.000

Sample 2000 (T)

Age

F.W.

(2212)  (178) (1664) (4054)

07–14 .42 15–25 .46 26–49 .55 >50 .47 Input 0.80 Range: .13

%

N

79 1910 78  130 80 1310 79 3350 Signif = 0.027

(T) (2409)  (166) (1635) (4210)

The age pattern attested in the 1980s Sample suggests a boundary between the younger speakers (first and second age groups) and the older; speakers under 25 years of age exhibit higher FWs (.51 and .57) than the other two groups. From twenty-six years of age onwards, the innovative variant declines and the conservative variant – the null subject – is retained mainly in the fourth age group (> 50 years old). The results displayed for the Sample 2000 show that there is no change among younger speakers and the second and fourth age-groups, but the speakers between 26 and 49 years move the spread of overt subjects in Rio de Janeiro. Then, the most important difference attested during the analysed lapse of time concerns the shift of the borderlines between age cohorts. The pattern identified in apparent time for 1980 suggests a generational change, while the pattern for 2000 one suggests age-grading variation. However, both patterns could also be interpreted in the following way: it is the second age-group, composed by adolescents, that leads the change. This could help explain why in 1980 the 15–25 year group shows a high FW and keeps this tendency in 2000, when they correspond to the third group. Then we can conclude that the process of change proceeds in the second synchrony. The analysis of the 16 subjects re-contacted brings more evidence for the overt referential subject as a continuing change towards completion in BP. The results displayed in Table 7 show that 75% of the re-interviewed speakers exhibit in each synchrony very close relative weights for overt subjects. The subset of 6 speakers who have continued their formal education shows regular behaviour: all of them remain stable, mapping the tendency attested in the overalls of the trend study. This regularity confirms the apparent time hypothesis and shows the importance of the adolescents. In the subset of speakers who have not furthered their formal education there is less consistency, but we can observe that more than 50% (6 out of 10) remain stable (FWs underlined). Only two women, both over 50 years of age (FWs in italics) increase their rates of overt subjects (Mgl48 and Nad36). Two other women, Eve43 and Jos35 (FWs in normal font), take

Chapter 11.  Panel and trend studies 187



Table 7.  Individual results on overt (vs null) subjects in sample 1980 and sample 2000 Speakers who furthered their formal education Speaker Eri59 Adr57 Adr63 Fat23 San39 Leo38

Gender

1980 Sample

woman woman woman woman woman man

 

2000 Sample

 

Age

F. W.

%

Age

F. W.

%

 9 10 12 15 15 18

.40 .51 .40 .66 .50 .43

74 83 76 90 83 78

25 26 28 33 33 36

.43 .50 .43 .69 .51 .51

79 84 79 91 83 85

.68 .50 .75 .65 .46 .78 .32 .46 .34 .25

88 80 92 84 76 92 65 76 70 61

Speakers who did not further their formal education Jup06 Lei04 Dav42 Jvas26 Eve43 Mgl48 Jan03 Nad36 Jos35 Ago33 Input .82

woman woman man man woman woman man woman woman man  

18 25 31 32 42 52 56 57 59 60  

.62 .53 .70 .65 .56 .65 .38 .26 .53 .27  

84 83 89 85 82 87 69 56 79 64  

35 43 48 48 59 70 74 74 75 77 Input .83  

 

an opposite direction in relation to the other speakers and the trend-study results, decreasing their rate of overt subjects. To sum up, we can say that, in general, there is convergence between the community and the individuals. Further evidence of the progress of the change comes from the effect of animacy and syntactic structure. The comparison of the two samples shows the weakening of restrictions concerning non-animate referents and same reference (as opposed to switch reference) in embedded contexts, which favoured the resistance of null subjects in Sample 1980 but become less significant in Sample 2000. 6. Final remarks In this paper we have shown that the combination of evidence drawn from trend and panel studies is a powerful methodological tool to identify the direction of three variable phenomena in spoken Portuguese of Rio de Janeiro. The comparison of samples of the same community in two different periods of time shows that the incrementation of preposition PARA and the loss of null referential subjects

188 Maria da Conceição de Paiva and Maria Eugênia Lammoglia Duarte

are continuing and irreversible changes, which affect not only the community as a whole but also the speakers as individuals. On the other hand, we then have communal change phenomena in Rio de Janeiro: the suppression of the anterior glide [j] showed a more complex situation. In this case, the trend study shows a reversal of the change in particular structural contexts in Rio de Janeiro. Since this reversal is embedded in another process of change, its adoption by individuals is more irregular and can explain the non-convergence between apparent time predictions and real time evidence.

References Bailey, Guy. 2002. Real and apparent time. In J. K. Chambers, Peter Trudgill and Natalie SchillingEstes (eds.), The handbook of language variation and change, 312–332. Malden-Oxford: Blackwell. Baxter, Garet and William Croft. 2016. Modeling language change across the lifespan: Individual trajectories in community change. Language Variation and Change 28. 129–173. https://doi.org/10.1017/S0954394516000077 Berlinck, Rosane. 1996. The Portuguese dative. In William Van Belle and Willian Langendonck The dative: Descriptive studies, 119–151. Amsterdam: John Benjamins Publishing. Bisol, Leda. 1989. Os ditongos na perspectiva da fonologia atual. D.E.L.T.A 5 (2). 185–224, São Paulo: Pontifícia Universidade Católica de São Paulo. Bisol, Leda. 1994. Ditongos derivados. D.E.L.T.A 10, Special issue. 123–140, São Paulo: Pontifícia Universidade Católica de São Paulo. Chambers, J. 2002. Patterns of variation including change. In J. K. Chambers, Peter Trudgill and Natalie Schilling-Estes (eds.), The handbook of language variation and change, 349–37. Oxford: Blackwell Publishing. Duarte, Maria Eugênia. 2000. The loss of the avoid pronoun principle in Brazilian Portuguese. In Mary Kato and Esmeralda Negrão (eds.), Brazilian Portuguese and the null subject parameter, 17–36. Vervuert: Iberoamericana. Duarte, Maria Eugenia. 2003. A evolução na representação do sujeito pronominal em dois tempos. In Maria da Conceição Paiva and Maria Eugênia Duarte (eds.), Mudança lingüística em tempo real, 115–128. Rio de Janeiro: Contra Capa-Faperj. Freire, Gilson. 2000. Os clíticos de terceira pessoa e as estratégias para sua substituição na fala culta brasileira e lusitana. (Master’s Thesis, Federal University of Rio de Janeiro, 2000). Retrieved from http://www.letras.ufrj.br/posverna/mestrado/FreireGC.pdf, accessed September 09, 2017. Freire, Gilson. 2005. A realização do acusativo e do dativo anafóricos de terceira pessoa na escrita brasileira e lusitana. (Phd Diss, Federal University of Rio de Janeiro, 2005). Retrieved from http://www.letras.ufrj.br/posverna/doutorado/FreireGC.pdf, accessed October 10, 2017. Gomes, Christina A. 1999. Directionality in linguistic change and acquisition. Language Variation and Change 11 (2). 212–230. Gomes, Christina A. 2003. Variação e mudança na expressão do dativo no português brasileiro. In Maria da Conceição Paiva and Maria Eugênia Duarte (eds.), Mudança lingüística em tempo real, 181–196. Rio de Janeiro: Contra Capa-Faperj.



Chapter 11.  Panel and trend studies 189

Labov, William. 1963. The social motivation of a sound change. Word 19. 273–309. Labov, William. 1972. Sociolinguistic patterns. Philadelphia: University of Pennsylvania Press. Labov, William. 1994. Principles of linguistic change: Internal factors. Oxford: Blackwell. Leite, Yonne, Dinah Callou and João Moraes. 2003. Processos em curso no português do Brasil: a ditongação. In Dermeval da Hora and Gisela Collishonn (eds.), Teoria Linguística: fonologia e outros temas, 232–250. João Pessoa: Editora Universitária. Lopes, Célia. 2002. Vossa Mercê > você e Vuestra Merced > usted: o percurso evolutivo ibérico. Lingüística 14. 173–190. Lopes, Célia. 2003. A inserção de ‘a gente’ no quadro pronominal do português. Frankfurt-Madrid: Iberoamericana, vol. 18. Nevalainen, Terttu, Helena Raumolin-Brunberg and Heikki Mannila. (2011). The diffusion of language change in real time: Progressive and conservative individuals and the time depth of change. Language Variation and Change 23. 1–43. Oliveira e Silva, Giselle M. and Maria M. Marta Scherre. 1996. Padrões sociolingüísticos: análise de fenômenos variáveis do português falado na cidade do Rio de Janeiro. Rio de Janeiro: Tempo Brasileiro. Paiva, M. da Conceição de. 1996. Supressão das semivogais nos ditongos decrescentes. In Giselle Machline de Oliveira e Silva and Maria Marta Pereira Scherre (eds.), Padrões Socio­ lingüísticos: análise de fenômenos variáveis do português falado na cidade do Rio de Janeiro, 325–334. Rio de Janeiro: Tempo Brasileiro. Paiva, Maria da Conceição de. 2003. O percurso da monotongação de [ey]: observações no tempo real. In Maria de Conceição de Paiva and Maria Eugênia Duarte (eds.), Mudança lingüística em tempo real, 31–46. Rio de Janeiro: Contra Capa-Faperj. Paiva, Maria da Conceição and Duarte, M. Eugenia. 2003. Mudança lingüística em tempo real. Rio de Janeiro: Contra Capa-Faperj. Robinson, John, S., Helen Lawrence and Sally Tagliamonte. 2001. GOLDVARB 2001: A multivariate analysis application for Windows. Sankoff, Gillian. 2006. Age: Apparent time and real time. Elsevier encyclopedia of language and linguistics. Second Edition. Article Number, LALI: 01479. Sankoff, Gillian and Hélène Blondeau. 2007. Language change across the lifespan: /r/ in Montréal French. Language 83. 566–588. Sankoff, Gillian and Hélène Blondeau. 2008. Longitudinal change accross the lifespan: /r/ in Montréal French. Language 83 (3). 560–588. Tagliamonte, Sali. 2011. Variationist sociolinguistics: Change, observation, interpretation. Oxford: Wiley-Blackwell.

Chapter 12

Quotative variation in Bernese Swiss German Christa Schneider, Sarah Grossenbacher and David Britain University of Bern

Whilst there have been a multitude of variationist studies on the changing quotative system of English, especially the diffusion of innovative be like (e.g. Tagliamonte and Hudson 1999; Buchstaller and D’Arcy 2009), work on the quotative systems of other languages (see Buchstaller and Van Alphen 2012) has been largely descriptive (e.g. Guardamagna 2010), either anchored in discourse analytic approaches (e.g. Mazeland 2006), or examining one type of quotative (e.g. Foolen et al. 2006), with few examining the linguistic and social constraints that shape speakers’ preference for certain variants over others (see, however, Palacios Martínez 2014; Cheshire and Secova, 2018). Research on German has been relatively limited and largely tackles quotatives from a conversation analytic perspective (e.g. Golato 2000; Bagi 2006; Imo 2007; Mertzlufft 2014). Here, we conduct a variationist analysis of the quotation system of a variety of German, Bernese Swiss German (BSG). All tokens of quotatives were extracted from a corpus of recordings of conversations with 26 working-class young adults from the western parts of the city of Bern, in which at least a third of the population is not ethnically Swiss. We therefore have recordings from speakers of Swiss German, but also, for example, from Bosnian, Bangladeshi and Albanian backgrounds. The results suggest that the quotative system in contemporary multiethnolectal Bernese German is shaped by a range of social and linguistic constraints, such as the presence or absence of quotative marker so, presence or absence of a quotative verb, presence or absence of a subject, speaker gender and speaker ethnicity. Keywords: quantitative analysis, quotatives, Swiss German, multi-ethnolect

1. Introduction Quotatives have become one of the most intensively studied variable phenomena in English. Their investigation has expanded our understanding, methodologically and theoretically, of language variation and change at the nexus of grammar, discourse-pragmatics and lexis, enabling us, subsequent to the advent of variationist https://doi.org/10.1075/silv.22.12sch © 2019 John Benjamins Publishing Company

192 Christa Schneider, Sarah Grossenbacher and David Britain

methods, to track an innovation from its very early days to a point where, in some varieties, it has become the dominant form. Scholars have been able to identify a set of linguistic and social constraints which operate on the choice of quotative form (addressing Weinreich, Labov and Herzog’s 1968 embedding problem), as well as examine where in the grammar the innovative forms enter, and how they encroach further through the system, addressing the transition problem. The existence of a multitude of studies around the Anglophone world has further facilitated comparative sociolinguistic analysis, considering the diffusion, embedding and localisation of variability across a number of speech communities simultaneously. And finally, we have been able, thanks to the rapidity of change, to consider real-time as well as apparent time change, and witness the trajectory of that change across the lifespan. Importantly, and perhaps surprisingly, however, there are very few prominent variationist studies of the quotative systems of languages other than English: Palacios Martínez (2014) analysed the quotative system among Spanish teenagers and Cheshire and Secova (2018) looked at Parisian French. There is a multitude of qualitative studies, for example, examining discourse markers which serve as quotatives, and studies investigating quotatives from a conversational analytic perspective. However, these are not able to or do not set out to address, to the same extent, the embedding and transition questions central to the understanding of change, facilitate comparability across studies, or track change through lifespan. Quotative research, for the most part, then, in languages other than English, is not yet able to even begin to address the questions that research on English has been tackling. It is the aim of this paper, therefore, to present a quantitative analysis of quotatives in German – a language that has seen much qualitative but no quantitative investigation. Most studies on German quotatives are based on qualitative analyses of the use of the verbs (sagen, denken, fragen – ‘say, think, ask’) and the use of the German adverb so ((‘like’). Furthermore, most of these studies are set in multi-ethnic communities. The following section will provide a brief summary of these studies. In Section 3 the methodology will be outlined. The results and analysis will be presented in Section 4. These results will be discussed in Section 5, where some final conclusions are also drawn. 2. Quotatives in German Golato (2000) analysed the use of innovative German so and suggests that this adverb is the German equivalent to English be like. She found that it was only used in narratives and points out that it serves as an enactment of performance, since it also allows for the use of voiced sound effects, gestures and facial expressions.



Chapter 12.  Quotative variation in Bernese Swiss German 193

Furthermore, she also reports that German so is usually preceded by a subject (ich ‘I’, er ‘he’, sie ‘she’, wir ‘we’) and followed by a micropause. With regards to the use of German so, it has been suggested that it is the German equivalent to English so, such (e.g. Wiese 2011). It is unstressed but stands adjacent to the clausal element that carries the main sentence stress. Furthermore, German so, she argues, is semantically bleached and reduced to a pragmatic/ information-structural function as a marker of focus. Merzlufft (2014) analysed how teenage girls use quotative constructions in phone calls. She suggests that German so is a marker of youth language, and acts as a scenic marker of performance allowing mimesis. Furthermore, German so can be stressed or unstressed, but is close to the main focus of the sentence. In addition, so is often used without a verb. Two studies report the use of German so in multi-ethnic communities: In her analysis of Kiezdeutsch, spoken in a number of multicultural neighbourhoods of Berlin, Wiese (2011) found that German so is used as a comparison particle, intensity marker, hedging strategy, a type token article and a new focus particle. In sum, she proposes that so is a new grammatical form generally serving to mark focus. There has been, as far as we are aware, no systematic qualitative or quantitative research on quotatives in Swiss German. There are, furthermore, very few studies on ethnolectal Swiss German (cf. Tissot et al. 2011). Given the lack of quantitative research on quotatives in German and the entire absence of research on them in Swiss German, our aim was to conduct an initial variationist analysis of the Swiss German quotative system. 3. Methodology In order to analyse quotative variation in Swiss German, data were collected during group interviews with 26 speakers who were between 17 and 24 years old. They all grew up in a multi-ethnolectal community in the western periphery of the city of Bern. Two social variables, namely gender (13 female and 13 male speakers) and ethnicity (9 speakers with a migration background and 17 Swiss speakers), were considered. The following five types of quotatives were found in the data: (1) Quotatives with a verb and pronoun de hesch du gseit, ‘nenei’ (then you said ‘no, no’) (2) Quotatives with a verb, pronoun and the particle so si seit immer so, ‘ja nei hm’ (she always says [so], ‘yes no hm’) (3) Quotatives with a pronoun and the particle so när ig so, ‘ja’ (then I [so], ‘yes’)

194 Christa Schneider, Sarah Grossenbacher and David Britain

(4) Quotative with so and without a pronoun u när so, ‘ja chani mi bewärbä’ (then [so], ‘yes can I apply’) (5) Quotatives with no pronoun, no so, but only när (then) and additional particles när schlussändlech, ‘ah es stimmt gar nid’ (then finally, ‘ah it’s not true’).

Variation occurs across eight parameters: (1) Presence of verb: a. Verb: när seit si so, ‘kenni’ (then she says [so] ‘I know’) b. Absence of verb: när ig so, ‘nei’ (then I [so] ‘no’) (2) Verb choice: a. Säge (say): när seit si so, ‘kenni’ (then she says [so] ‘I know’) b. Dänke (think): mir hei natürlech so dänkt, ‘hei Leo, ah chum, hallo, geits guet’ (we just [so] thought, ‘hey Leo, come on, hello, are you ok’) c. Frage (ask): när hei si mi gfragt, ‘how much is this’ (and then they asked me, ‘how much is this) d. Finge (find): wüu är eifach so fingt ‘ja, boah ey, mache immer die ganzi zit z gliiche’ (because he simply finds, ‘well yes, I’m doing the same thing the whole time) e. Cho (come): wie si si de zu dir cho, ‘hesch du dini huusufgabe gmacht’ (how they then to you come, ‘did you do your homework’) f. Sii (be): är isch so, ‘ciao ciao’ (he is [so], ‘ciao, ciao’) g. Others: dr Phippu meint so ‘dä söt das gar nid intressiere’ (Philip thinks [so], you should not be interested in that) (3) Tense of verb: a. Present: är isch so, ‘what the fuck’ (he is like, ‘what the fuck’) b. Perfect: i ha se gfragt, si het gseit, ‘oh’ (I asked her, she said, ‘oh’) (4) Presence of subject: a. 1. Pers. Sing.: und ig so, ‘mhm, ja genau’ (and I [so], ‘mhm, yes sure’) b. 2. Pers. Sing.: u när hesch du drum no so gseit, ‘ah, mami lueg’ (and then you said [so], ‘ah, Mom look’) c. 3. Pers. Sing.: u när si, ‘nid’ (and then she [no verb], ‘not’ d. 1. Pers. Pl.: mir so, ‘okay’ (we [so] [no verb], ‘okay’) e. 3. Pers. Pl.: när het si gseit, ‘aso itz chasch schüsch i lift ine ga’ (then they said, ‘well now you get in the lift) f. Personal names: und när seit mer d Claudia ‘ja weisch si geit itz i ds Griechisch’ (then Claudia tells me ‘yeah, you know, she takes Greek classes now’) g. Absence of subject: u da inegange, ‘oh entschuldigung’ (and [no subject] went in, ‘oh sorry’)



Chapter 12.  Quotative variation in Bernese Swiss German 195

(5) Presence of so: a. So: mir so, ‘okay’ (we [so] [no verb], ‘okay’) b. Absence of so: u da inegange, ‘oh entschuldigung’ (and [no subject] went in, ‘oh sorry’) (6) Position of so: a. So + verb: u när hani so gfragt: ‘ja’ (and then I [so] asked, ‘yes’) b. Subj. + so: när ig so, ‘nei’ (then I [so] [no verb], ‘no’) c. Verb + so: und säge so, ‘du, iss itz’ (and [no subject] say [so], ‘you, eat now’) (7) Use of adverb: a. De (then): u de so ‘ah ds si dini kollege’ (and then [so] [no verb], ‘oh, these are your friends’) b. Immer (always): si seit immer so, ‘ja nei, hm’ (she always says [so], ‘yeah no, hm’) c. När (then): när mir, ‘ja super, was mache mer itz’ (then we [no verb], ‘yeah great, what are we doing now’) d. No (also): ja äbä und i ha no gseit, ‘brrr’ (yes well, and I also said, ‘brrr’) e. Absence of adverb: mir so, ‘okay’ (we [so] [no verb], ‘okay’) (8) Type of quote a. Normal (voice does not change): när ig so ‘ja’ (and then I [so], ‘yes’) b. Mimesis (voice changes, imitation): Drillmaster si die wo säge ‘du go fuckin bastard go down go!’ (Drillmasters are those who say ‘you go fuckin’ bastard go down go!’) c. Non-lexicalised sound: und i ha no gseit ‘brrr’ (and I even said ‘brr’)

One difficulty in conducting a variationist analysis of Swiss German quotatives is that it is hard to establish what the dependent variable should be, since the use of quotative verbs, quotative markers such as so and subject pronouns are all optional in the construction of a quotative in Swiss German. We therefore conducted a range of analyses of the dataset, but ultimately focussed on the deployment of so, given its surface similarity with English ‘like’. Both descriptive and inferential statistics (logistic regression analysis using Rbrul) were used. The results are presented in the following section. 4. Results In total 399 quotative tokens were found in our data. We begin by presenting descriptive statistics which outline the distribution of the tokens in our dataset. Section 4.2. looks more closely at the use of so in a logistic regression analysis to determine which social or linguistic variables are most likely to influence its presence.

196 Christa Schneider, Sarah Grossenbacher and David Britain

4.1

Descriptive statistics

Firstly, the presence of a verb is not necessary in a Bernese Swiss German quotative (see Figure 1). In total, 171 tokens (42.9%) were without a verb, and 228 (57.1%) were with a verb – see examples in (1) above. 250

Number of tokens

200

150

100

50

0

Present

Absent

Figure 1.  Number of quotative tokens where a verb was absent or present

Furthermore, the use of subjects is not compulsory – see the examples in (4) above. Although the majority of tokens makes use of a subject (339 tokens, 85%), there were still 60 tokens without a subject (see Figure 2). 400 350

Number of tokens

300 250 200 150 100 50 0

Present

Figure 2.  Presence of subjects

Absent

Chapter 12.  Quotative variation in Bernese Swiss German 197



The use of so is also variable – see examples in (5) above. As can be seen in Figure 3, so was included in 211 of the 399 tokens (52.9%). 250

Number of tokens

200

150

100

50

0

Present

Absent

Figure 3.  Presence of so

If so is used, it can occur in a number of positions (examples in (6) above). It is mostly used following a subject (99 tokens, 24.8%). In addition there were 38 tokens (9.8%) with so in other positions: these mainly include Adverb+ so, or where so stands alone. There were 37 tokens (9.3%) where so followed a verb, and 37 (9.3%) 120 tokens where so preceded it.

Number of tokens

100

80

60

40

20

0

other

Figure 4.  Position of so

so + Subj.

so + V.

V. + so

Subj. + so

198 Christa Schneider, Sarah Grossenbacher and David Britain

4.2

Linguistic and social constraints on the use of so

Given the fact that so has received some attention in studies on German quotatives (cf. Golato 2000; Wiese 2011; Merzlufft 2014), and its equivalent ‘like’ has been the driving motivation behind the vast majority of recent studies of quotatives in English, we focus here on presence or absence of so within the quotative as the dependent variable for further statistical analysis. In order to find out which social and linguistic constraints influence the choice of so, a logistic regression analysis was carried out using Rbrul (Johnson 2009). Speaker was used as a random intercept. It was found that the type of quote (p  .05) do not significantly influence the use of so. A significant tendency was found regarding the presence or absence of verbs (p  % of known and used loanwords

Figure 3.  Distribution of the acquaintance with and use of Romance loanwords by different age groups in Betina. The analysis is based on the sample of 100 lexical items. (Škevin 2012: 175)

4.3. Zadar The results based on the Zadar study show that the speakers use Romance loanwords in 33,46% of cases, a standard variant in 51,15% of the cases, and in 15% both (Figure 4). When they use both variants, they sometimes claim to use the Romance variant only when speaking to their parents or grandparents. The more abrupt decline in the use of Romance vocabulary in Zadar can be interpreted in different ways. Being an urban centre of Northern Dalmatia, Zadar is more open to immigrants from other parts of the country and particularly from the Štokavian hinterland unlike the other two sites of research. Moreover, an early shift from Čakavian to Neoštokavian back in the 1960s (Brozović 1975/76: 53), which Betina and Vis have been undergoing more intensely only recently, was also enhanced by education and state media which promote the Standard Croatian variety.

212 Ivana Škevin Rajko and Lucija Šimičić

60 51

50 40

34

(%) 30 20

15

10 0

Use of dialectal variants of Romance origin

Use of standard variants of (not necessarily) Slavic origin

Use of both variants

Figure 4.  Use of dialectal variants by young adult speakers in Zadar (age 20–45). The analysis is based on the sample of 25 lexical items

5. Discussion and conclusions The small-scale comparative analysis presented here shows differential resistance of the Romance lexical stratum in four ČVs. In this study we focused exclusively on the analysis of nouns characteristic of semantic domains pertaining to everyday life in which Romance influence is prominent with regard to other (non-Čakavian and especially non-coastal) Croatian dialects. Such items are generally less prone to erosion, and have proven to represent a stable aspect of South Čakavian identity in the insular varieties under study (like katrida ‘a chair’, ponistra ‘a window’, kušin ‘a pillow’, travesa ‘an apron’, škure ‘blinds’) (Škevin 2010, Šimičić 2011). However, these more than other parts of the lexicon seem to have been affected in the urban context of Zadar due to its exposure to immigration accompanied by the importation of alternative linguistic models and linguistic innovations. This is why the Romance lexemes for such concepts in Zadar have already been replaced by their SC equivalents (for instance, instead of ponistra ‘a window’ the speakers use SC prozor). The reasons for Romance lexical base maintenance in Betina, Komiža and in Vis may lie in the fact that rural, especially insular communities, are more closed or less attractive to immigrants (with the exception of the summer periods), which makes them more linguistically conservative. If the shift occurs, it is more common among younger speakers and in those varieties that have more direct and frequent connections with urban centres on the coast (in Betina more than on the island of Vis, and in Vis more than in Komiža). Moreover, in the insular varieties even in those domains in which Romance elements are relatively well preserved, there is a tendency towards levelling for the concepts that are less generic, less frequently used (such as bruncin ‘a type of a cooking bowl’), and those that are not common



Chapter 13.  Dialect levelling or shift 213

(any more) in RD (such as kamara ‘a room’, kočeta ‘a bed’ or skulor ‘a pupil’), which may be ascribed to their greater salience. Although it is not always easy to define the scope and the exact meaning of linguistic salience, and despite the fact that salience is primarily a psychological rather than a linguistic term and hence may vary (what’s salient for younger urban speakers may not be salient for the older ones), it is sometimes the Romance origin of a word (e.g. prsura ‘a frying pan’ in Betina and in Zadar, as opposed to SC tava) that makes it salient in the sense suggested here. When Romance-origin lexemes are present in both local Čakavian varieties and in RD, salience seems to be a characteristic of more locally specific items as they are generally less visible, less comprehensible in a regional context, and also often exposed to stereotyping (see Section 2). There are many instances when Betina speakers, for instance, lexically converge to RD by abandoning more local variants not known or used regionally (regional ponistrica instead of local herijada ‘a small window’, or regional kanotjera instead of local gučica ‘an undeshirt’). While the potential salience of lexical variants of Romance origin, though under-investigated, is still very likely, we suggest that it is a factor which plays a role in speakers’ accommodation which can lead to dialect levelling and ultimately to dialect shift. On the other hand, a variant’s origin and its markedness can cause positive stereotyping in local contexts so that the use of such items can become crucial in geographical and linguistic identification of a speaker (e.g. bublija in Betina ‘a type of cake baked for Easter’). Another cause of dialect levelling lies in the fact that some lexical variants are already considered obsolete by younger speakers even though the older speakers still use them. In the case of both salient and obsolete lexical variants, younger speakers tend not to use them even when they know their meaning. In Betina, it is the case of dontrina (‘a religious education’), kočeta (‘a bed’) or kamara (‘a bedroom’), in Zadar, it is the case of travesa/traverša (‘an apron’) or prsura (‘a frying pan’), while in Vis, such are the terms skulor (‘a pupil’), koćeta (‘a bed’), etc. Areal and structural criteria were thus shown to play only a limited role in salience of particular features (Auer, Barden, and Grosskopf 1998), and subjective factors such as perceptual distance, attitudes and prestige may be equally important in accounting for levelling effects. This small-scale analysis shows that lexical levelling taking place in an urban centre in many cases leads to convergence towards SC, but structural advergence towards nonstandard regional variety (RD in our case) is more common in local, rural and insular, varieties. In this way the two rural varieties are more conservative, while the Zadar urban variety is extremely innovative in terms of the extent of the observed lexical levelling. Moreover, the psychological factors seem to play a greater role in Betina, Komiža and in Vis where marked, salient or obsolete lexical

214 Ivana Škevin Rajko and Lucija Šimičić

items tend to disappear, whereas the shift is more likely to occur in Zadar because of immigration as a key external factor. This finding thus reveals a relatively recent tendency as historically the centres of linguistic innovation and stability on a South Čakavian-Štokavian continuum showed opposite patterns with insular varieties being much more innovative compared to their coastal South Čakavian counterparts. This preliminary conclusion, however, calls for more detailed analysis in terms of dialect levelling and the role of salience in such processes present at other levels of linguistic structure, which would be beyond the scope of this paper.

References Auer, Peter, Birgit Barden, and Beate Grosskopf. 1998. “Subjective and Objective Parameters Determining ‘Salience’ in Long-term Dialect Accommodation”. Journal of Sociolingustics 2 (2): 163–187. Auer, Peter, Frans Hinskens, and Paul Kerswill. 2005. “The Study of Dialect Convergence and Divergence: Conceptual and Methodological Considerations”. In Dialect change. Con­ vergence and Divergence in European Languages, ed. by Auer, Peter, Frans Hinskens, and Paul Kerswill, 1–51. Cambridge: Cambridge University Press. Brozović, Dalibor. 1975/76. “O suvremenoj zadarskoj miksoglotiji i o njezinim društvenopovijesnim i lingvističkim pretpostavkama [On contemporary Zadar mixoglotty and its socio-historical and linguistic determinants].” Radovi Filozofskog fakulteta u Zadru 14–15: 49–63. Brozović, Dalibor, and Pavle Ivić. 1988. Jezik srpskohrvatski/hrvatskosrpski, hrvatski ili srpski [The Serbo-Croatian / Croato-Serbian, Croatian or Serbian]. Zagreb: Jugoslavenski leksikografski zavod. Elšík, Viktor. 2009. “Loanwords in Selice Romani, an Indo-Aryan Language of Slovakia.” In Loan­words in the World’s Languages: A Comparative Handbook, ed. by Martin Haspelmath, 260–303. Berlin: Mouton de Gruyter. Field, Fredric W. 2002. Linguistic Borrowing in Bilingual Contexts. Amsterdam: John Benjamins Publishing. Haspelmath, Martin. 2008. “Loanword Typology: Steps Toward a Systematic Cross-Linguistic Study of Lexical Borrowability.” In Aspects of Language Contact: New Theoretical, Methodological and Empirical Findings with Special Focus on Romancisation Processes, ed. by Thomas Stolz, Dik Bakker, and Rosa Salas Palomo, 43–62. Berlin: Mouton de Gruyter. Haspelmath, Martin. 2009. “Lexical Borrowing: Concepts and Issues“. In Loanwords in the World’s Languages: A Comparative Handbook, ed. by Martin Haspelmath, and Uri Tadmor, 35–54. Berlin: Mouton de Gruyter. Kerswill, Paul, and Ann Williams. 2002. “‘Salience’ as an Explanatory Factor in Language Change: Evidence from Dialect Levelling in Urban England.” In Language Change: The Inter­play of Internal, External and Extra-Linguistic Factors. ed. by Mari C. Jones, and Edith Esch, 81–110. Berlin: Mouton de Gruyter.



Chapter 13.  Dialect levelling or shift 215

Kerswill, Paul. 2003. “Koineization and Accommodation”. In The Handbook of Language Variation and Change, ed. by J. K. Chambers, Peter Trudgill, and Natalie Schilling-Estes, 669–702. Oxford: Blackwell Lisac, Josip. 1999. “Leksička norma i hrvatska narječja [The lexical norm and the Croatian groups of dialects]”. In Hrvatski govori, filolozi, pisci [Croatian spoken varieties, philologists, writers], ed. by Josip Lisac, 166–178. Zagreb: Matica Hrvatska. Prifti, Elton. 2016. “Osservazioni su una storia romanza dei contatti linguistici interadriatici [Observations on a Romance history of interadriatic linguistic contacts]”. In Relazioni linguistiche. Strutture, rapporti, genealogie [Linguistic relationships. Structures, relations and genealogies], ed. by Martin, Becker, and Ludwig, Fesenmeier, 185–210. Frankfurt: Peter Lang. Šimičić, Lucija. 2005. Analiza temeljnog i kulturnog leksika viških govora i njihov odnos prema drugim istraživanim govorima istočnog Jadrana [Analysis of basic and cultural vocabulary of the varieties on the Island of Vis and their relationship to the previously studied dialects on the eastern Adriatic coast]. Zagreb: University of Zagreb [Unpublished manuscript: pre-doctoral thesis]. Šimičić, Lucija. 2011. Identifikacijski procesi i jezične promjene na otoku Visu [Identification processes and language change on the Island of Vis]. Zagreb: University of Zagreb [unpublished PhD thesis]. Šimičić, Lucija, Peter Houtzagers, Anita Sujoldžić, and John Nerbonne. 2013. “Diatopic Patterning of Croatian Varieties in the Adriatic Region”. Journal of Slavic Linguistics, 21(2): 259–301. Šimičić, Lucija, and Anita Sujoldžić. 2009. “Istraživanje temeljnog i kulturnog leksika u naseljima otoka Visa: prostorni i generacijski aspekti [The study of basic and cultural vocabulary on the Island of Vis: spatial and generational aspects]”. In Destinacije čežnje, lokacije samoće: Uvidi u kulturu i razvojne mogućnosti hrvatskih otoka [Desired Destinations, Lonely Locations: Insights in Culture and Developement of the Croatian Islands], ed. by Ines Prica, and Željka Jelavić, 189–202. Zagreb: Hrvatsko etnološko društvo. Škevin, Ivana. 2010. Etimološka i leksikološka obradba posuđenica romanskog podrijetla u govoru mjesta Betina na otoku Murteru [Etymological and lexicological analysis of Romance loanwords in the variety of Betina on the island of Murter]. Zadar: University of Zadar [unpublished PhD thesis]. Škevin, Ivana. 2012. “Između arhaičnog (romanskog) i standardnog (hrvatskog) jezičnog elementa: koineizacija otočnih varijeteta [Between the archaic(Romance) and standard (Croatian) linguistic element: koineization of island varieties]”. In Zbornik radova Aktualna istraživanja u primjenjenoj lingvistici [Contemporary research in applied linguistics: The proceedings], ed. by Leonard, Pon, Vladimir Karabalić, and Sanja Cimer, 171–184. Osijek: HDPL. Škevin, Ivana. 2016. “Dialect levelling and changes in semiotic space”. In The Future of Dialects. Selected Papers from Methods in Dialectology XV. Language Variation 1, ed. by Côté, MarieHélène, Remco Knooihuizen, and John Nerbonne, 281–304. Berlin: Language Science Press. Trudgill, Peter. 1986. Dialects in Contact. Oxford: Blackwell. Žanić, Ivo. 2017. Jezična republika. Hrvatski jezik, Zagreb, Split i popularna glazba [The linguistic republic. Croatian language, Zagreb, Split and popular music]. Zagreb: Jesenski i Turk.

Chapter 14

Complementing in another language Prosody and code-switching Jonathan Steuck and Rena Torres Cacoullos Pennsylvania State University

In English-Spanish code-switching, the main and complement clause boundary is a site of variable equivalence between languages. Whereas the complementiser is always present in Spanish, in English it is only sometimes present, giving rise to a quantitative word string mismatch at this juncture. Comparisons with monolingual benchmarks reveal no grammatical convergence of the contact varieties in finite complementation patterns. Rather, prosody provides a solution to variable equivalence. Whereas main and complement clauses tend to be prosodically integrated by occurring in the same Intonation Unit in unilingual speech, the opposite is true when there is code-switching at the clause boundary. Prosodic distancing of the two languages at junctures of variable equivalence is thus a bilingual strategy for code-switching between separate grammars. Keywords: prosody, complementation, code-switching, equivalence, convergence, bilingualism, English, Spanish, New Mexico, Intonation Unit

1. Code-switching and equivalence Mixing is a descriptor widely applied to bilingual speech. Mixing of grammars is also implied by blended labels such as Spanglish or Türkendeutsch. Yet grammatical convergence is not inherent to bilingualism. Bilinguals’ grammars are interconnected, as evidenced in structural priming across languages (Gries and Kootstra 2017), but still separate, as evidenced in their aligning with their respective monolingual benchmarks, while differing from each other (Torres Cacoullos and Travis 2018). The key methodological tool is the identification of grammatical (dis)similarities by using quantitative diagnostics in speech corpora.1

1. We gratefully acknowledge National Science Foundation grants 1019112/1019122 to Rena Torres Cacoullos and Catherine Travis and 1624966 to Torres Cacoullos and Shana Poplack. https://doi.org/10.1075/silv.22.14ste © 2019 John Benjamins Publishing Company

218 Jonathan Steuck and Rena Torres Cacoullos

Code-switching between languages is illustrated in (1). Examples are from the New Mexico Spanish-English Bilingual corpus (cf. Torres Cacoullos and Travis 2018, Chapters 2 and 3); italics and roman type respectively indicate speech originally produced in Spanish and English. (1) Miguel: … hay veces que quiero poner una .. Spanish word in there. ‘… there are times that I want to put a .. Spanish word in there.’  [04, 1: 11: 19–1: 11: 24]2

The Equivalence Constraint states that bilinguals tend to avoid code-switching at points of word order conflicts between the two languages. Word order equivalence was the major finding of Poplack (1980), who reported that fewer than 1% (11/1,835) of code-switches occurred at points where the word orders were different. The key insight of the Equivalence Constraint is that the concatenated strings constituting code-switches internally follow the grammar of their respective language, but that at the points at which languages are switched the two grammars are “equivalent” (Poplack 1980; Sankoff 1998). For example, in (1), where Miguel (a pseudonym) begins in Spanish and ends in English, each string adheres to the respective monolingual norms. On the one hand, the Spanish string follows the Spanish ‘there are times that’ rather than the English ‘there are times when’ (which is preferred over there are times that; a search in the Corpus of Contemporary American English [Davies 2008–] yielded frequencies of 856 versus 49). On the other hand, the English string displays adjective noun order rather than the Spanish preference for postnominal adjectives (e.g. Delbecque 1990). However, at the juncture between languages, there is word string equivalence between the two languages, since the determiner precedes the noun in both Spanish and English (Sankoff and Poplack 1981: 17). Here we focus on the boundary of main and complement clauses, illustrated in (2) for English and (3) for Spanish. In (2), the English complementiser that is absent, indicated with a Ø between the main clause [MC] and the complement clause [CC]. The complementiser is in fact absent most of the time in English (see Shank, Plevoets and Van Bogaert [2016] for a summary of that/zero alternation). Though rates of complementiser presence vary according to register, they range approximately between 10% and 20% in corpora of spoken English (e.g. Thompson and Mulac 1991; Torres Cacoullos and Walker 2009). In Spanish, in contrast, the complementiser que is virtually always present (Silva-Corvalán 1994: 137).

2. Within brackets following examples are the recording number and the beginning-ending time stamps of the lines reproduced. Transcription protocols follow Du Bois et al. (1993).



Chapter 14.  Complementing in another language 219

(2) Pedro: … I thought [MC] Ø it was a pretty big town back then. [CC]  [10, 35: 55–35: 57] c14-sq3 (3) Alfredo: yo pensé [MC] que estaba muy alto. [CC] ‘I thought [MC] that it was very high. [CC]’

[31, 52: 11–52: 12]

For English-Spanish bilinguals, the juncture between MC and CC would be a site of “variable equivalence” (Torres Cacoullos and Poplack 2016). Sites of variable equivalence display a quantitative rather than a qualitative mismatch – where the languages differ only sometimes, due to independent, but inherently variable, processes in one or both of the languages. The sometimes-applying mismatch between English and Spanish in the presence of the complementizer raises the problem of variable equivalence for code-switching at this juncture. In this paper, we offer evidence for prosodic separation as a bilingual strategy for dealing with cross-language mismatches found at the juncture of two languages. 2. A community-based bilingual speech corpus To address the prosody of code-switching, we visit northern New Mexico, where code-switching is “the appropriate code for the Hispano community” (Gonzales 1999: 29). Spanish has been spoken in this region longer than English (Bills and Vigil 2008: 29–47). With U.S. statehood in 1912, however, Spanish was relentlessly displaced in the schools, being relegated to a foreign language subject. Today the local variety is endangered, due to stigmatisation and shift to English. Still, of U.S. states, New Mexico has the highest Hispanic or Latino population as a percent of the total, at 47%, followed by California and Texas, each at 38% (United States Census Bureau 2015). It also has the lowest proportion of first-generation immigrants (foreign born), at about one-fifth of the New Mexico Hispanic population, compared with approximately one third in Texas and nearly half in Florida (United States Census Bureau 2015). In northern counties, the percent of Hispanics is as high as 80% and the proportion of non-US born is as low as 5%. Since these demographics imply limited contact with monolingual Spanish, this bilingual community is a prime candidate for language mixing. What kind of data offer insights on patterns of language use within the speech community? According to Labov (1989: 2), “Word lists and formal elicitations are the primary sources of confusion in our descriptions of language.” The New Mexico Spanish-English Bilingual (NMSEB) corpus is a sociolinguistic-style corpus of spontaneously produced bilingual speech, recording 40 northern Hispanic New Mexicans, minimally third-generation, who meet the criterion of regular use of both languages (Torres Cacoullos and Travis 2018: 13–56). Participants are not

220 Jonathan Steuck and Rena Torres Cacoullos

asked to code-switch nor about their code-switching. Instead, the technique is to elicit narratives of personal experience, where speakers shift towards the vernacular, the mode of everyday speech, because they, not the university-affiliated interviewer, are the authority (Labov 1984: 32). Testimony to the bilingualism of the corpus is the distribution of [Main Clause + Complement Clause] instances according to language. In Figure 1 we see that the proportions of English-only, as in (2), and Spanish-only, as in (3), tokens are nearly equal, at 41% (467/1,133) and 43% (484/1,133), respectively. 50 40 (%)

41

43

30 20 10 0

English

Spanish

9

7

CS

Lone

Figure 1.  Language of [Main Clause + Complement Clause] (NMSEB, N = 1,133)

Cases involving multiword code-switching (CS) constitute 9% (107/1,133) of [Main Clause + Complement Clause] tokens. In (4), code-switching is at the boundary between the clauses (@ represents one syllable of laughter). In (5) code-switching is within one of the clauses, here between the verb and an adverbial expression in the complement clause. Lone items, or single word insertions (as in me acuerdo que era una purple ‘I remember that it was a purple one’ [06, 18: 28–18: 29]), which add up to another 7% (75/1,133), are not counted with multiword (unambiguous) instances of code-switching (cf. Poplack 2018). c14-s2q (4) Code-switching at [MC + CC] boundary Dolores: me dijeron que, [MC] I was gonna run the two mile? [CC] ‘they told me that, [MC] I was gonna run the two mile?’ [CC]

[22, 11: 08–11: 10]

c14-s2q5 (5) Code-switching within MC or CC, but not at boundary Dora: … se me hace [MC] que era four years ago. [CC] ‘…I think [MC] that it was four years ago.’ [CC]  [20, 30: 10–30: 11]

When speakers switch from one language to the other at the boundary between main and complement clause, they do so in both directions at roughly the same rate. Speakers switch from Spanish main clauses to English complement clauses, as in (4) above, about 59% of the time, and the reverse, as in (6), 41% of the time in aggregate, as shown in Figure 2. The lack of a conspicuous directionality of

Chapter 14.  Complementing in another language 221



code-switching is further testimony to the bilingualism of these speakers. The even distribution of the languages and lack of switching directionality are consistent with the locally unpredictable use of code-switching as a general discourse mode for the community. (6) Fabiola: …(0.8) did Nancy tell you [MC] que había venido a ver a la grandma? [CC] ‘…(0.8) did Nancy tell you [MC] that she had come to see grandma?’ [CC]  [09, 43: 43–43: 46] 59

60 50 40

41

(%) 30 20 10 0

English[MC]-to-Spanish[CC]

Spanish[MC]-to-English[CC]

Figure 2.  Direction of code-switching at [Main Clause + Complement Clause] boundary (NMSEB, N = 63)

3. Prosodic and syntactic relationships We do not usually speak in the well-formed sentences privileged by much syntactic analysis. Intonation Units are recognisable units of conversation (Chafe 1994: 53– 70). The Intonation Unit (IU) is “a stretch of speech uttered under a single, coherent intonation contour” (Du Bois et al. 1993: 47). Each IU appears on a distinct line in the transcription followed by punctuation representing its transitional continuity (Du Bois et al. 1993: 53). For example, in (7) the first IU has continuing intonation, indicated by a comma; the second IU ends in a final intonation contour, marked by a period. (7) Sandra: I think he just was giving us time to think, you know.

[03, 18: 49–18: 51]

Transcription of the speech stream into IUs is done perceptually by trained linguists (Du Bois et al. 1993).3 Nevertheless, the wave form and F0 trace in Figure 3 illustrate acoustic features transcribers attend to. For example, we can see the higher pitch at 3. For each hour of speech, at least 50 were required for transcription of the NMSEB corpus (Torres Cacoullos and Travis 2018: 46–51).

222 Jonathan Steuck and Rena Torres Cacoullos

F0 (Hz)

0 300

0.5

1.0

1.5

250 200 150 I

think he

just

was giving

us

time

to

think,

I think he just was giving us time to think,

Figure 3.  Acoustic properties of the Intonation Unit (IU) *Figure created in Praat (Boersma and Weenink 2018)

the beginning, gradually dropping over the course of the IU to the end, as well as the slower rate of speech at the end of the IU (compare the two instances of think). There is generally a tighter syntactic relationship between words in the same IU than between words positioned in different IUs (Croft 1995: 849–864). The syntactic relationship between main and complement clause is realized prosodically in the tendency for the complement clause verb to occur in the same IU as its main clause verb. This is true for finite clausal complements in conversational English and Spanish data, at rates of 78% (660/844) and 68% (224/328) respectively (Steuck 2016: 81), as shown in Figure 4 (left panel). The datasets consist of all declarative finite verbs with finite clausal complements. Conventionalised formulae such as you know or collocations such as I think that are prosodically independent or that occur as parentheticals were excluded (see Steuck [2016: 77–80] for exclusion protocols). So, most of the time main and complement clauses are prosodically integrated in a single IU. When are they spread over two or more IUs? In Table 1, contexts hypothesised to influence the prosody of complementation are operationalised as factors in a multivariate analysis; their relative contribution to the probability of the absence of prosodic integration is given as a factor weight between 0 and 1 (Sankoff et al. 2005). The higher the number, the greater the probability of the absence of prosodic integration in the corresponding context. The second column gives the number of observations. In both languages (monolingual benchmarks appear in the leftmost columns of Table 1), the absence of prosodic integration is favoured in the context of

Chapter 14.  Complementing in another language 223



90

Monolingual English benchmark (N = 844)

80

(%)

90

78

70

60

60 (%)

40 30 20

90

Two (or more) IUs

0

Same IU

Monolingual Spanish benchmark (N = 328)

90

Same IU

Bilinguals’ Spanish (N = 484) (NMSEB)

70

64

60

50

(%) 32

50 40 20

10

10 Two (or more) IUs

36

30

20 0

Two (or more) IUs

80 68

70

30

22

10

60 40

40 20

80

(%)

50 30

22

10 0

78

80

70 50

Bilinguals’ English (N = 467) (NMSEB)

Same IU

0

Two (or more) IUs

Same IU

Figure 4.  Prosodic integration of [Main Clause + Complement Clause]: Proportion of occurrences in same IU versus two (or more) IUs in monolingual benchmarks and bilinguals’ varieties *Monolingual benchmarks from Steuck (2016): Santa Barbara Corpus of Spoken American English (SBCSAE) (Du Bois et al. 2000–2005); Corpus of Conversational Colombian Spanish (CCCS) (cf. Travis 2005: 9–25)

intervening material, such as adverbials, other clauses, or pauses.4 In English, the presence of the complementiser, itself a kind of intervening material, also favours the absence of prosodic integration (in Spanish, the complementiser is invariably 4. Pauses are represented by dots in the transcription; a number within parentheses indicates a timed pause of 0.7 seconds or longer, as in Example (6). Though pauses often coincide with IU boundaries, they are not a required feature (Chafe 1994: 5: 58–60; Croft 1995: 840). IU-internal pauses, as in (6), are notably less frequent; in a NMSEB sample, 33% (550/1662) of IUs have pauses marked initially, but only 5% (81/1662) have internal pauses (Steuck 2018: 86–88).

224 Jonathan Steuck and Rena Torres Cacoullos

present). Grammatical person and form of the subject also contribute to speaker choices: the absence of prosodic integration is more likely with non-first person main clause subjects and with lexical complement clause subjects. These factors indicate the role of syntactic distancing, as well as that of particular constructions (e.g., I think, yo creo ‘I believe’) and considerations of information status or heaviness of constituents (Croft 1995: 850–864; Steuck 2016). Table 1.  Independent variable rule analyses of the contribution of linguistic factors to the absence of prosodic integration of [Main Clause + Complement Clause] EnglishMonolingual

SpanishMonolingual

EnglishBilingual

SpanishBilingual

Average

22%

184/844 32%

104/328 22%

101/467 36%

Corrected mean

.22

.27

.16

.32

Intervening material Present Absent

.92 .34

126/183  58/661

.90 .30

  70/91 34/237

.94 .37

  62/80 .89  39/387 .28

Complementizer Present Absent

.81 .45

  60/107 124/737

  NA

   328  0

.71 .42

55/126 46/341

Person main clause subject 3rd or 2nd person .62 1st person .46

 77/222 107/622

.61 .41

61/150 43/178

[.57] [.48]

33/104 [.50] 68/363 [.50]

106/310  66/174

Form compl clause subject Lexical [.59] [.46] Other

 70/246 [.65] 114/598 [.46]

32/69 72/259

[.58] [.48]

 25/90 76/376

 48/104 123/377

  NA

.65 .46

172/484

118/150  54/334     484  0

* Factor weights in variable rule analysis (Sankoff 1988) indicate the strength of constraints relative to the corrected mean. Monolingual benchmark analyses from Steuck (2016)

What, then, is the prosodic profile of bilinguals’ code-switched [Main Clause + Complement Clause] occurrences? 4. Variable equivalence and English-Spanish complementation At junctures of variable equivalence the word strings of the two languages in contact are equivalent only sometimes. As we have seen, such is the boundary of main and complement clauses: whereas in Spanish the complementiser que is always present, in English complementiser that is variably present. Bilinguals might circumvent the mismatch between the languages by mixing them, that is, through grammatical convergence (cf. Gumperz and Wilson 1971; Backus 2005). In other words, they



Chapter 14.  Complementing in another language 225

might make the contact varieties more similar to each other than their non-contact (monolingual) counterparts are. They do not do so. Evidence against grammatical mixing is that bilinguals’ English-only and Spanish-only [Main Clause + Complement Clause] occurrences adhere to the same prosodic pattern as their respective monolingual benchmarks. The rates of integration of main and complement clauses in the same IU are 78% (N = 467) in bilinguals’ English, but 64% (N = 484) in their Spanish (Figure 4, right panel). At the same time, the direction of effect of factors contributing to the prosody of complementation is consistent across monolingual and bilingual varieties of the two languages (Table 1, right-hand columns). (The difference in main-clause subject person in bilinguals’ New Mexican Spanish is due to the use of the impersonal expression se me hace ‘it seems to me’ [as in (5)], absent in the Colombian Spanish benchmark; see Figure 6.) A second piece of evidence countering convergence is that the distribution of main-clause verbs is strikingly similar in the bilingual and the monolingual varieties. In English, for both bilingual and monolingual varieties, think, know, guess and say account for two-thirds of the data (Figure 5, top panel). In Spanish, the most frequent complement-taking verbs are creer ‘believe’, decir ‘tell’, saber ‘know’, pensar ‘think’ and, in the New Mexican data, se me hace ‘it seems to me’ and acordarse ‘remember’ (Figure 5, bottom panel). A third piece of evidence that bilinguals are adhering to separate grammars is that complementiser that is mostly absent in bilinguals’ English (appearing at a rate of 27%, 126/467), as in the English monolingual benchmark, while complementiser que is never absent, as in the Spanish benchmark (Table 1). In sum, at least as concerns finite complementation, when these bilinguals speak English, they speak as monolingual English speakers do. When they speak Spanish, they do so like other Spanish speakers. So, when switching at the main and complement clause boundary, bilinguals confront a word string mismatch between their languages, because of the language-internal structural variability in English. What strategy do they use, then, to deal with such a site of variable equivalence? One hypothesis is that bilinguals employ prosodic distancing of code-switching boundaries (Torres Cacoullos and Poplack 2016). As we have seen, main and complement clauses tend to be prosodically integrated, that is, they tend to occur in the same prosodic unit (Figure 4, above). Therefore, the prediction is that, in the presence of code-switching at the boundary between main and complement clauses, absence of prosodic integration will be more frequent, as compared with Spanish-only and English-only baselines.

226 Jonathan Steuck and Rena Torres Cacoullos

100 90 80 70 60 (%) 50 40

Other; 32

Other; 28

say; 9 guess; 6 know; 10

say; 8 guess; 10 know; 6

think; 43

think; 48

Monolingual English

Bilinguals’ English (NMSEB)

Other; 30

Other; 34

30 20 10 0

100 90 80 70 60 (%) 50 40 30

pensar; 8 saber; 13 decir; 24

0

se me hace; 12 decir; 26

20 10

acordarse; 8 saber; 9

creer; 25 Monolingual Spanish

creer; 11 Bilinguals’ Spanish (NMSEB)

Figure 5.  Frequency of complement-taking lexical types, in monolingual benchmarks and bilinguals’ varieties

*Monolingual benchmarks: English from Torres Cacoullos and Walker (2009: 20); Spanish calculated from the Corpus of Conversational Colombian Spanish (CCCS) (cf. Travis 2005: 9–25)

5. Code-switching through prosodic distancing of the boundary between main and complement clause Code-switching in bilingual complementation may be at the boundary between main and complement clause – as in Examples (4) and (6) – or within one of the clauses, as in (5). Are main and complement clauses prosodically integrated at the same rate when code-switching occurs at the clause boundary – a site of variable equivalence – as when it occurs within clauses at loci of equivalence? If code-switching entails a general processing cost, the absence of prosodic integration should be equally as high, compared with monolingual baselines. If, on the

Chapter 14.  Complementing in another language 227



other hand, it is equivalence that matters and bilinguals use prosody to deal with variable equivalence, the absence of prosodic integration should be more frequent with code-switching at the clause boundary. Figure 6 (left panel) shows that in the presence of within-clause code-switching, the main and complement clause tend to occur in the same IU, at a rate of 64% (28/44), nearly the same as in the monolingual benchmarks and – notably – as in bilinguals’ own non-code-switched speech (Figure 4). But with code-switching at the clause boundary, we see the opposite tendency, namely not to prosodically integrate the clauses, at a rate of 60% (38/63) (Figure 6, right panel). Code–switching within main or complement clause (N = 44) (NMSEB)

(%)

Code–switching at boundary of main and complement clause (N = 63) (NMSEB)

70

70

60

60

50

50

40

40

(%)

30

30

20

20

10

10

0

Two (or more) IUs

Same IU

0

Two (or more) IUs

Same IU

Figure 6.  Prosodic integration of [Main Clause + Complement Clause]: Proportion of occurrences in same IU versus two (or more) IUs, in the presence of code-switching (*p