Prosodic and Rhythmic Aspects of L2 Acquisition : The Case of Italian [1 ed.] 9781443846851, 9781443842471

Citation preview

Prosodic and Rhythmic Aspects of L2 Acquisition

Prosodic and Rhythmic Aspects of L2 Acquisition: The Case of Italian

Edited by

Anna De Meo and Massimo Pettorino

Prosodic and Rhythmic Aspects of L2 Acquisition: The Case of Italian, Edited by Anna De Meo and Massimo Pettorino This book first published 2012 Cambridge Scholars Publishing 12 Back Chapman Street, Newcastle upon Tyne, NE6 2XX, UK British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Copyright © 2012 by Anna De Meo and Massimo Pettorino and contributors All rights for this book reserved. No part of this book may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the copyright owner. ISBN (10): 1-4438-4247-8, ISBN (13): 978-1-4438-4247-1

TABLE OF CONTENTS

Introduction ................................................................................................. 1 Interlanguage Prosody Anna De Meo and Massimo Pettorino Part I: Linking Practice and Theory Chapter One............................................................................................... 15 Prosodic Features in Native and Non-Native Speech Segmentation Patrizia Sorianello Chapter Two .............................................................................................. 37 Rhythmic-Prosodic Aspects in the Acquisition of Italian as a Second Language: A Teaching Experience with Adult Foreign Learners in Catania Milena Romano Chapter Three ............................................................................................ 59 Pragmatic Markedness in L2 Italian: Some Data on the Morph-Syntax/ Intonation Interface in Moroccan Arabic Speakers’ Interlanguage Laura Mori Chapter Four.............................................................................................. 83 “And Nevertheless They Do Speak Italian”: Prosodic Explorations in Speech of Deaf Immigrants Elisa Pellegrino and Valeria Caruso Chapter Five ............................................................................................ 103 About “Ethnics” Italian’s Prosodies: Data and Observations Domenico Russo Part II: The Role of L1 Chapter Six .............................................................................................. 115 Geminate Timing in the Speech of Estonian L2 Learners of Italian Chiara Celata and Lidia Costamagna

vi

Table of Contents

Chapter Seven.......................................................................................... 137 A Rhythmic-Prosodic Analysis of Italian L1 and L2 Luciano Romito and Andrea Tarasi Chapter Eight........................................................................................... 153 Requesting in Italian as a Foreign Language: Pragmatic and Prosodic Features Iolanda Alfano, Claudia Crocco and Renata Savy Chapter Nine............................................................................................ 177 Yes-No Questions in Belgian Dutch (L1) and Italian (FL) Jens Baele Part III: Prosody and Pragmatics Chapter Ten ............................................................................................. 199 Rhythm in Narration and Description in L2 Italian Marta Maffia, Anna De Meo and Massimo Pettorino Chapter Eleven ........................................................................................ 213 Foreign Accent and Persuasiveness: Native and Non-Native Voices in a Radio Spot Marilisa Vitale, Anna De Meo and Massimo Pettorino Chapter Twelve ....................................................................................... 223 The Strategic Use of Pauses and Intonation in L2 Italian Argumentative Speech: The Case of an Intercultural Debating between Chinese and Italians Luisa Salvati Chapter Thirteen...................................................................................... 237 Prosodic-Pragmatic Interface in L2 Italian Learners’ Repetitions Anna De Marco and Patrizia Sorianello Part IV: Technology and Prosody Recognition Chapter Fourteen ..................................................................................... 263 Automatic Prosodic Comparison between Model and Imitation Sentences in a Second Language Teaching Computerized Environment Philippe Martin

Prosodic and Rhythmic Aspects of L2 Acquisition: The Case of Italian

vii

Chapter Fifteen ........................................................................................ 277 Text-to-Speech Synthesis: Phonetic Mapping Applied to Italian as Target Language Pierluigi Salza, Albert De Groot and Enrico Zovato References ............................................................................................... 297 Contributors............................................................................................. 331

INTRODUCTION INTERLANGUAGE PROSODY ANNA DE MEO AND MASSIMO PETTORINO UNIVERSITY OF NAPLES “L’ORIENTALE”, ITALY

1. Prosodic aspects of communication The word “prosody” refers to some features of speech, traditionally defined as “suprasegmental”, which include stress, rhythm, tone and intonation. The non-discrete nature of the prosodic elements and the assumption about their purely expressive function has overshadowed the importance of these aspects of the language for a long time. However, since the 1980s, studies in experimental phonetics and applied linguistics (sociolinguistics, discourse analysis, pragmatics, systemic-functional linguistics) have led to their re-evaluation, as studying language in context has allowed linguists to point out the distinctive and contrastive functions of prosody and to highlight its contribution to the syntactic, semantic and pragmatic characterization of the message in spoken communication.1 One of the most reassessed prosodic features has been intonation, i.e. the melodic movement of an utterance depending on the control over the glottal mechanism activity. For this trait of human language, acquired in the very early months of life,2 elements of interlinguistic variability, both structural and functional, and universal characteristics were identified.3 1 Dorothy Chun, Debra M. Hardison and Martha C. Pennington, “Technologies for prosody in context: Past and future of L2 research and practice”, in Phonology and second language acquisition, eds. G. J. Edwards Hansen and Mary L. Zampini (Amsterdam: John Benjamins, 2008), 323-46; Julia Hirschberg, “Pragmatics and Intonation,” in The Handbook of Pragmatics, eds. Laurence R. Horn and Gregory Ward (Oxford: Blackwell Publishing, 2006), 515-37. 2 According to Sarah Hawkins, “The most global properties of the baby’s to-benative language are prosodic and some of these may even be learned at or before birth. Rhythm and pitch patterns can be heard in utero, since they are what is left in the signal when the mother’s voice reaches the uterus, low-pass filtered through

2

Introduction

In recent years studies in both language acquisition and language education have included prosody, especially intonation,4 focusing attention on the L1-L2 relationship and on the problem of prosodic transfer.5 Even so, rhythmic, tonal, intonational and accent related systems, all elements characterizing language-dependent prosody, are still rarely taught6 and yet they are the basis of human communication. her body”. Sarah Hawkins, “Auditory Capacities and Phonological Development: Animal, Baby and Foreign Listeners”, in The Acoustics of Speech Communication, ed. James M. Pickett (Needham Heights, MA: Allyn and Bacon, 1999), 183-98. 3 Jacqueline Vaissière, “Phonetic explanations for cross-linguistic prosodic similarities”, Phonetica 52 (1995):123-30; Jennifer Fitzpatrick, “On intonational typology”, Methodological issues in language typology, ed. Peter Siemund, Sprachtypologie und Universalien Forschung 53 (2000): 88-96; Alan Cruttenden, “Falls and rises: meanings and universals”, Journal of Linguistics 17 (1981): 77-91. 4 Katsura Aoyama and Susan G. Guion, “Prosody in second language acquisition: Acoustic Analyses of duration and F0 range”, in Language experience in second language speech learning: In honor of James Emil Flege, eds. Oke-Schwen Bohn and Murray J. Munro (Amsterdam: John Benjamins, 2007), 282-97; Dorothy M. Chun, Debra M. Hardison and Martha C. Pennington, “Technologies for prosody in context: Past and future of L2 research and practice”, in Phonology and second language acquisition, eds. Jette G. Edwards Hansen and Mary L. Zampini, (Amsterdam: John Benjamins, 2008), 323-46; Pavel Trofimovich and Wendy Baker, “Learning second-language suprasegmentals: Effect of L2 experience on prosody and fluency characteristics of L2 speech”, Studies in Second Language Acquisition 28 (2006): 1-30; Ineke Mennen, “Bi-directional interference in the intonation of Dutch speakers of Greek”, Journal of Phonetics 32 (2004), 543-563; Jürgen Trouvain and Ulrike Gute, eds., Non-native Prosody. Phonetic Description and Teaching Practice (Berlin/New York: Mouton de Gruyter, 2007); Wolfgang Grosser, “Aspects of intonation L2 acquisition”, in Current Issues in European Second Language Acquisition Research, eds. Bernhard Kettemann and Wilfried Wieden (Tübingen: Gunter Narr Verlag, 1993), 81-94. 5 Kees De Bot, “The transfer of intonation and the missing database”, in Crosslinguistic Influence in Second Language Acquisition, eds. Eric Kellerman and Michael Sharwood Smith (New York: Pergamon Press, 1986), 110-19; Laurent Rasier and Philippe Hiligsmann, “Prosodic transfer. Theoretical and Methodological Issues”, Nouveaux cahiers de linguistique Française 28 (2007), 41-66; Empar Devís Herraiz, “La prosodia nell’interferenza tra L1 e L2: il caso delle interrogative polari tra veneti e catalani”, Estudios de fonética experimental 16 (2007), 119-46. 6 Dorothy M. Chun, Discourse intonation in L2: From theory and research to practice (Amsterdam: John Benjamins, 2002); Jürgen Trouvain and Ulrike Gut, eds., Non-native prosody: phonetic description and teaching practice, (Berlin: Mouton de Gruyter, 2007); Jesús Romero-Trillo, Pragmatics and Prosody in English Language Teaching (Dordrecht: Springer, 2012).

Interlanguage Prosody

3

2. Prosody and L2 Italian Although second language acquisition studies show a growing sensitivity towards prosodic competence, which is an element considered crucial for the identification of a non-native pronunciation7 and for message understanding,8 much remains to be studied with reference to the Italian language. Italian is not among the most widely spoken languages in the world in terms of native speakers, but it is so for the number of language courses devoted to foreigners,9 who start studying it for both cultural and professional reasons (Foreign Language - FL).10 The large-scale immigration experienced in Italy since the 1980s also posed the problem of spontaneous or mixed learning of Italian as a second language (L2) by subjects who vary in terms for age, mother tongue, level of literacy, level of integration in the arrival social context, migration project duration, motivation etc.11 An ever increasing number of people use Italian as a 7 Janet Anderson-Hsieh, Ruth Johnson and Kenneth Koehler, “The Relationship Between Native Speaker Judgments of Nonnative Pronunciation and Deviance in Segmentals, Prosody, and Syllable Structure”, Language Learning 42/4 (1992): 529–55. 8 Tracey M. Derwing, Murray J. Munro, “Accent, comprehensibility and intelligibility: Evidence from four L1s”, Studies in Second Language Acquisition 19 (1997), 1-16; Tracey M. Derwing, Murray J. Munro, Grace Elaine Wiebe, “Pronunciation instruction for ‘fossilized’ learners: Can it help?”, Applied Language Learning 8 (1997), 217-35. 9 Data published in the survey Italiano 2000, sponsored by the Italian Ministry of Foreign Affairs, show that Italian language is the fifth most studied foreign language in the world. Cf. Tullio De Mauro and Massimo Vedovelli, Italiano 2000. I pubblici e le motivazioni dell’italiano diffuso fra stranieri (Roma: Bulzoni, 2003). 10 See also Massimo Vedovelli, L’italiano degli stranieri. Storia, attualità e prospettive (Roma: Carocci, 2002). 11 Monica Barni and Andrea Villarini, La questione della lingua per gli immigrati stranieri: insegnare, valutare e certificare l’italiano L2 (Milano: Franco Angeli, 2001); Beatrice Iori and Sergio Neri, eds., L’italiano e le altre lingue, Apprendimento della seconda lingua e bilinguismo dei bambini e dei ragazzi immigrati (Milano: Franco Angeli, 2005); Fernanda Minuz, Italiano L2 e alfabetizzazione in età adulta (Roma: Carocci, 2005); Antonella Benucci, ed., L’italiano libera-mente. L’insegnamento dell’italiano a stranieri in carcere (Perugia: Guerra, 2008); Stefano Rastelli, Italiano di cinesi, italiano per cinesi. Dalla prospettiva della didattica acquisizionale (Perugia: Guerra, 2010); Massimo Vedovelli, Stefania Massara and Anna Giacalone Ramat, Lingue e culture in contatto. L’italiano come L2 per gli arabofoni (Milano: Franco Angeli, 2001);

4

Introduction

vehicle of communication, from foreign students and tourists to jobseeking immigrants (from Africa and Asia, as well as Latin America and eastern Europe), political and environmental refugees, unaccompanied minors, children joining their immigrant parents, and undocumented migrants who have escaped extreme poverty, or ethnic strife and political oppression in their own countries. This volume therefore stems from the need to deeply study problems regarding the suprasegmental level of L2 Italian. It addresses, in particular, the variability related to oral discourse genres, the communicative contexts, the pragmatic functions, the native/non-native interaction, the relation between interlanguages and varieties of L1 Italian (dialects, regional varieties, etc.), the influence of mother tongues, the speaker’s specific characteristics, the teachability of prosody, the sociolinguistic issues related to the foreign accent perception, and the impact of technology.

3. Between Practice and Theory This book opens with an overview on the possible approaches to the study of the rhythmic-prosodic skills acquisition in a second language (Linking Practice and Theory), carried out through the application of diverse analytical tools and models. Patrizia Sorianello proposes an acoustic-instrumental study on synthesized and spontaneous speech, with the aim of investigating the production and perception of prosodic boundaries. The complexity of doing this research is caused by various factors, such as temporal rules (pauses and variation in speech tempo) and melodic criteria, e.g. intonational variations, pitch range and pitch reset. The first step in this experimental survey, carried out on the natural speech of L1 Italian speaking subjects and Polish L2 Italian speakers (level A2), all living in Bari, south east Italy, shows the profound difference in language production existing between native and non-native speakers: variations in the number of syllables of Tonal Units (TU) and in the speech rate, and a different use of pauses, both silent or filled, were observed. In particular, the non-native speech, using short TUs, long silent pauses and non-verbal items, appear prosodically disorganized. A perceptual test then conducted on L1 and L2 Italian synthesized speech (tone flattened, devoid of silent and filled pauses, and modified combining both manipulations) verified Rossella Bozzone Costa, Luisa Fumagalli and Ada Valentini, eds., Apprendere l’italiano da lingue lontane: prospettiva linguistica, pragmatica, educativa (Perugia: Guerra, 2012).

Interlanguage Prosody

5

the ability of native listeners to identify the position of intonational boundaries. In this regard, listeners used different perceptual strategies, which varied according to the native or the non-native nature of the listened stimuli. However, listeners’ perception was not only influenced by prosody, but also by the overall fluency level. Milena Romano chooses a sociolinguistic perspective. Taking into account the complex situation of contemporary Italian, characterized by neostandard and regional varieties beside the standard language, the author shows the process of suprasegmental skills and metalinguistic awareness development related to the diatopic, diaphasic, diamesic and diastratic variability, through a questionnaire and a perceptual test based on television speech. The observations, conducted in Catania, Sicily, compared young adult immigrants and foreign students coming from different European and non-European countries. The results show that the use of television texts for teaching purposes allows delocalisation of spoken Italian and overcoming of the regional variant as to rhythmicprosodic and phonetic traits. In Laura Mori’s study, prosodic analysis becomes a tool for interpreting the functional value of morpho-syntactic structures; she investigates the use of the right-dislocated construction ce + lo. In particular, she focuses on this cataphoric pronoun cluster when, followed by the inflected verb form avere (to have), it signals in advance the sentence comment, with regard to the speech production of Moroccan immigrants, spontaneous learners of Italian. Averci, sociolinguistic variant of avere frequently attested in popular informal registers of spoken Italian, is preceded by the accusative clitic lo in the interlanguage of the considered subjects. This construction appears as an anomalous reduplication in discourse, since the right-dislocated clitic phonological word just anticipates the comment. Mori considers such use of language as structurally and pragmatically marked, and since marked sentence structures usually exploit prosodic prominence configurations, she instrumentally analysed the morphosyntax/intonation interface for evaluating the informative value of the construction under investigation. Results confirm that, as assumed, nonnative speakers use the clitic cluster as a verbal-agreement marker and as a tool for emphasizing the informative relevance of the linguistic unit in focus position as well. Pellegrino and Caruso present some preliminary “prosodic explorations” in the speech of deaf immigrants. After a comprehensive overview of the medical literature devoted to the analysis of the prosodic and vocal features of deaf people’s oral productions, the authors synthesize the results of three task-based tests, used to examine the L2 Italian prosodic

6

Introduction

competence of the deaf participants. If, on the one hand, they underline the impact of the hearing deficit on the speaking skills of deaf immigrants, on the other hand they prove the deaf person’s ability to manage some prosodic features and their pragmatic awareness that communication develops through informative chunks. Moreover, this study also highlights the social role that linguists could play in helping those who are in need of a specific language education, thus completing the assistance of medical care given to these hearing-impaired speakers. In the end, Pellegrino and Caruso suggest that further research on spontaneous speech production is needed, and solicit a major involvement of linguists in this kind of studies. A completely different approach is taken by Domenico Russo, who argues against the possibility of comparing the characteristics of native speech with those of a non-native one, which he labels as “ethnic”. According to the author, the extreme variability of L1 and L2 speeches, together with the effect of an “unconscious camouflage”, revealed in all the productions, makes “neutral” comparisons between native and L2 corpora almost impossible, in terms of voice quality at both the segmental and the prosodic level.

4. Prosodic Transfer from L1 to L2 The second part of the book, The Role of L1, emphasizes the relationship between the mother tongue and L2, and investigates the presence of transfer in the prosody interlanguage development. Celata and Costamagna analyze the interlanguage prosodic tuning of native Estonian speakers, who are acquiring the Italian singletongeminates consonant contrast. This study is based on the spectro-acoustic analysis of read and repeated speech, of both isolated words and short sentences in Italian, and particularly investigates the temporal relationship between vocalic segments and adjacent consonants. The singletongeminate consonantal opposition occurs in both involved languages, even if it is rendered in different ways: it is binary in Italian (short and long) and ternary in Estonian (short, long and extra-long). In the interlanguage of the Estonian participants, an adaptation to the Italian phonological system was observed as regards the stop consonants, while for continuous consonants there rather seems to be a transfer effect from the L1, especially as far as rhotic sounds are concerned. Romito and Tarasi analyze a corpus consisting of spontaneous and read speech produced by adult immigrants, speaking Chinese, Romanian, Polish and Albanian as their L1s. All the subjects involved in this study come from Calabria, southern Italy, and learned Italian as a second

Interlanguage Prosody

7

language in a non-guided learning context. Thanks to acoustic, perceptual, musical and automatic speech recognition-related analysis, the authors state that speakers tend to keep their L1 rhythmic and segmental patterns in different ways, which varies depending on the L1 typological characteristics. Furthermore, the authors propose an “Italian index”, that is, a modality of visual representation of the distances between the different segmental and suprasegmental features existing amid L1 Italian and the various interlanguages. Major differences are those related to the varieties produced by Chinese and Polish speakers, while Albanian speech productions are the closest to the L1 Italian, being characterized by dissimilarities limited to duration of vocalic and consonant elements within stressed syllables speech productions. In the studies of Alfano, Crocco, Savy, and Baele, the attention shifts towards Belgian learners of Italian as foreign language (B1, B2, C1 levels of the CEFR), all speaking Dutch as L1,. Spontaneous and semispontaneous dialogues, elicited through the “spot the difference” technique, compose the corpus used for two surveys. Alfano, Crocco and Savy’s purpose is to analyze the pragmatic and prosodic realization of L1 and FL Italian requests, and deals especially with the query_y move realization for direct requests, which is the most frequently used pragmatic move in each of the three examined groups. Analysis outcome shows that the general competence in FL does not significantly correlate with the pragmatic skills improvement, whereas a high proficiency level matches with a more appropriate use of intonation. Baele focuses on yes-no questions, divided into two groups: conversational moves, or queries, and confirmation requests for given information, or check moves. The pragmatic difference under analysis is expressed in L1 Italian trough prosody. This study, based on the analysis of comparable read speech corpora uttered in Belgian Dutch and Italian, showed a prosodic transfer from L1 to the FL concerning boundary tones and tonal categories, which appears stronger in B1 and B2 learners.

5. Prosody as a Pragmatic Phenomenon The third part of the book, Prosody and Pragmatics, illustrates the role that prosody plays in the interpretation of pragmatic meaning in nativenon-native interaction, in the realization of L2 Italian narrations, description and argumentation, and its influence on repetitions and message persuasiveness. Maffia, De Meo and Pettorino compare narration and description in the semi-spontaneous speech of two native Italian speakers and two adult

8

Introduction

immigrants having different L1s—Swahili and Wolof—with different competence levels in L2 Italian (respectively A2 and B1 of the CEFR). Subjects were asked to perform three linguistic tasks, characterized by different rhythmic-prosodic features and having similar trends in L1 and L2: the description of a picture; a visually guided storytelling, requiring the production of a short narration elicited by a comic strip; and the narration of an emotionally involving personal experience. Results showed that in L2 Italian the description is defined by a lower speech rate and longer and more frequent silent pauses, maybe due to a greater difficulty encountered by the speakers in the discourse planning. These problems are, however, minimized in the narration, thanks to the temporal sequence of events, which is used as a guiding pattern for the structuring of the utterance. The personal and more involving narration was presented with a more fluent speech than in the other two tasks, and with a higher percentage of articulated sequences, especially for the learner having the lowest level of competence in L2 Italian. Generally speaking, a relationship between the degree of variability of the statement and the level of L2 proficiency was found: for the lowest level, sentences are more diversified in terms of rhythm and prosody, depending on the task typology. Luisa Salvati’s aim was to investigate the prosody production in the argumentative speech of Chinese learners of L2 Italian, involved in an intercultural debating with Italian speakers. This study points out two important considerations concerning the relationship between prosody and second language acquisition. The first one deals with intercultural debating as a valid tool for collecting L2 spontaneous speech corpora, and as an effective task affecting L2 learners’ cognitive and communicative skills. The second one concerns the importance of the spectro-acoustic analysis, since its results not only show the relevant role played by prosody in L2 Italian spontaneous speech, but its socio-phonetic and pragma-phonetic interpretation also allows observation how non-native speakers use and manage voice in order to persuade. Vitale, De Meo, Pettorino deal with non-native Italian speakers’ persuasiveness from an experimental perspective. Through an acousticperceptive analysis carried out on radio speech of French, Russian, Chinese and Italian speakers, the study aims at identifying the acoustic perceptual correlates of persuasiveness, in order to verify whether a persuasive communication depends on the degree of foreign accent, or whether it is rather related to language-independent variables. Spectroacoustic analysis and perceptive test outcome point out that foreignaccented Italian speakers seem to be limited in the achievement of a

Interlanguage Prosody

9

persuasive speech and that, by contrast, for both native and non-native speakers a less varied and more flattened tone favours persuasiveness. Repetition is under investigation in the study by De Marco and Sorianello, who offer a general overview of the different functions of this communicative strategy. The analysis was carried out on a corpus consisting of image descriptions and semi-structured interviews between native and non-native speakers of Italian; the L2 subjects were four Turkish and two Spanish speakers, all having an A1 level of competence in Italian, and one American with a B2 level of the CEFR. Thanks to the phonetic markedness of repetitions in terms of duration, F0 contour, intensity changes, pauses and glottalizations, prosody has been used to disambiguate some of the functions emerging in the corpus. “Self-repair”, “stalling”, “expanding” and “control over the form or meaning” were the most frequent kinds of repetitions noticed in the corpus, and the various functions they cover are sometimes different for A1 and B2 speakers. The B2 participant uses “self-repair” for organizing discourse, just as native speakers do, while A1 subjects use the same strategy as a self-correction of lapses. In most cases, planning problems are revealed by “stalling” repetitions, while, by contrast, conversation is favoured by “expansions”; the main function of the “control over the form or meaning” is to ask for agreement about the accuracy and adequacy of the repeated element.

6. Technology meets L2 Prosody In the fourth part of this volume, Technology and Prosody Recognition, technology meets prosody in two specific areas: i.e. second and foreign language teaching, and speech synthesis systems that allow a computer to transform a written text in natural language, artificially reproducing the human voice. Martin introduces the theme of technology in the service of prosody teaching. As a matter of fact, prosody is a fundamental factor in message decoding, and it should therefore undoubtely also have a primary role in the teaching of an L2 oral skills. Since the 1960s, several computer software packages have been developed with the purpose of helping learners to acquire proper oral skills. However, according to Martin, these tools are limited regarding the phonological insight in their implementations. All these types of software, which usually provide a graphical display of teacher and learner’s F0 curves, allow learners to imitate a model and then graphically compare it with their own performances. For various reasons this approach has not achieved satisfactory results, e.g. the lack of phonological and rhythmic information useful to evaluate the curve

10

Introduction

appropriateness. In order to address the latter issue, Martin presents a new set of functions implemented in the WinPitch software that allows learners’ productions to be stylized on a model utterance, in terms of F0 curve and rhythmic structure, through automatic morphing. Learners can thus imitate their resynthesized voices and be provided with instruction on the elements requiring corrections and adjustments. In the study by Salza, de Groot and Zovato, technology is applied to speech synthesis. In particular, they present the improvement of the Loquendo Text-to-Speech (LTTS ®) phonetic mapping technique, used for mixed language text reading obtained through the modification of the default output of the Automatic Phonetic Mapping (APM) module, which runs in the multilingual synthesis system. This module allows each L1 voice to speak any L2 language provided by the system, using L1 phonetic system as the basis for choosing L2 phonemes: for this purpose, a distance function exploiting weightings assigned to each articulatory-phonetic feature is used. However, since the acoustic-prosodic return of the APM concatenations is not always satisfactory, rules for choosing L1 candidate phonemes to be mapped onto the L2 phoneme has been defined in an Experimental Phonetic Mapping (EPM) prototype. More specifically, L1 phonemes are selected taking into account L1 and L2 phonetic systems, the frequency of occurrence of the candidate phoneme in the L1 speech database, and the similarity evaluation on a phonetic sequence longer than the single phoneme. Examples of the EMP use, having fluent, intelligible and plausible pronunciation of L2 Italian, are provided for the French-toItalian and German-to-Italian mappings.

7. Conclusions All recent studies in spoken communication emphasize the importance of the suprasegmental level in conversational interaction, and the prosodic competence may be undoubtedly considered more crucial than the lexical and the morpho-syntactic ones in conveying the proper interpretation of utterance meaning and function. The contribution of this field of study to second language acquisition research is crucial, as prosody is not merely related to the big divide between native and non-native speakers, in terms of their accuracy and efficacy in production, but it also deals with the input domain control. Prosody may influence in a decisive way the input perception and decoding, with negative effects on L2 language acquisition. This volume is the first study introducing the prosodic development of Italian as a second/foreign language, and it covers a wide range of topics, offering different standpoints from which the complexity and variety of

Interlanguage Prosody

11

the issue become clear. Studies have been conducted at various levels of analysis, from the micro-prosody of the syllable to the macro-prosody of the pitch contour of the sentence or the rhythm of the entire text, and there are reflections and discussions with reference to different research fields, i.e. language acquisition and teaching, pragmatics, sociolinguistics, computational linguistics, conversation analysis, communication disorders. In this volume the editors decided to deal with the prosodic interlanguage phenomena from a multiplicity of methodological approaches and theoretical perspectives. All the work, albeit using different research methods and having their strengths and weaknesses, shares the common goal of contributing to the understanding of interlanguage prosody and to the development of communicative proficiency in a second language.

PART I LINKING PRACTICE AND THEORY

CHAPTER ONE PROSODIC FEATURES IN NATIVE AND NON-NATIVE SPEECH SEGMENTATION PATRIZIA SORIANELLO UNIVERSITY OF BARI, ITALY

1. Introduction In communicative dynamism, intonational phrasing is a reliable device which allows listeners to encode the information structure and the planning of spoken discourse. The prosodic boundary’s identification is of great importance to the detection of coherent stretch of speech (for both semantic and syntactic levels) with a global melodic cohesion. There has been a considerable number of studies dealing with the phonetic description of prosodic boundaries in recent decades1. The topic was discussed from many points of view, in production as well as in perceptual perspective, employing either natural speech or a synthesized one. However, in other

1

Cf. Lise Menn and Suzanne Boyce, “Fundamental frequency and discourse structure”, Language and Speech 25 (1982): 341-83; Jan Roelof de Pijper and Angelien A. Sanderman, “On the perceptual strength of prosodic boundaries and its relation to suprasegmental cues”, Journal of Acoustical Society of America 96 (1994): 2037-47; Stephen Schuetze-Coburn and Elizabeth G. Weber, “Units of intonation in discourse: a comparison of acoustic and auditory analyses”, Language and Speech 34 (1991): 207-37; Mark Swerts, “Prosodic features at discourse boundaries of different strength”, Journal of Acoustical Society of America 101 (1997): 514-21; Mark Swerts and Ronald Geluykens, “The prosody of information units in spontaneous monologue”, Phonetica 50 (1993): 189-96; Mark Swerts and Ronald Geluykens, “Prosody as a marker of information flow in spoken discourse”, Language and Speech 37/1 (1994): 21-43.

16

Chapter One

studies, the prosodic boundaries’ detection was accomplished using sophisticated automatic algorithms or prediction models.2 Several studies investigated the acoustic correlates of prosodic phrase boundaries. At present, it’s widely accepted that the realization of a prosodic boundary is affected at least by two groups of factors. On the one hand there are temporal criteria including pauses and variations in speech tempo,3 while on the other are melodic criteria such as intonational variations, pitch range and pitch reset. In addition, a minor role is generally attributed to changes in voice quality (i.e. laryngealization) and energy variations (i.e. loudness). Previous research has widely shown that prepausal lengthening, local changes in F0 contours, and pause duration are used for demarcative purposes: a boundary tone is melodically realized with a noticeable pitch variation aligned with the edges of tone units (henceforth TU), through the lengthening of the TU final syllables, or also by the insertion of a pause. These features co-occur in different ways, signalling boundaries of various prosodic strength. It should be remarked that, to a certain extent, these acoustic parameters are used in a similar way across language to mark prosodic breaks.4 Intonational boundaries need to be studied combining, in close synergy, acoustic measurements and perceptual evaluations, albeit that the agreement values are not always concordant, as repeatedly demonstrated by the ratings of agreement obtained during the annotation of prosodic markers accomplished by using both acoustic cues and listeners’ judgements, by expert and non-expert researchers. 2

See among others, Franco Cutugno and Leandro D’Anna, “Segmenting the speech chains into tone units human behaviour versus automatic processes”, in Maria Josep Solé, Daniel Recasens and Joaquín Romero (eds.), Proceedings of the 15th International Congress of Phonetic Sciences, Barcelona, Spain 3-9 August 2003, Barcelona 2003, Universitat Autonòma de Barcelona, 1233-36; Agnieszka Wagner, “Acoustic cues for automatic determination of phrasing”, in Proceedings of the 5th International Conference on Speech Prosody, Chicago, Illinois May1114, Illinois 2010, University of Illinois Press, accessed June 2012, http://speechprosody2010.illinois.edu/program.php/papers/100196.pdf. 3 See Li-Chiung Yang, “Duration and pauses as phrasal and boundary marking indicators in speech”, in Maria Josep Solé, Daniel Recasens and Joaquín Romero (eds.), Proceedings of the 15th International Congress of Phonetic Sciences, Barcelona, Spain 3-9 August 2003, Barcelona 2003, Universitat Autonòma de Barcelona, 1791-94. 4 Cf. Alan Cruttenden, “Falls and rises: meaning and universals”, Journal of Linguistics 17 (1981): 77-91; Michael D. Tyler and Anne Cutler, “Cross- language differences in cue use for speech segmentation”, Journal of Acoustical Society of America 126 (2009): 367-76.

Prosodic Features in Native and Non-Native Speech Segmentation

17

Prosody has an important role in language learning and use. Despite the great quantity of studies devoted to this topic in first language (conventionally designed as L1), non-native use of prosodic cues in speech segmentation received far less scholarly interest. In more recent years, however, prosody in second language (L2) has received growing attention in more experimental directions. One line of research attempted to define the perception of foreign accents5 and, more generally, the mastery of prosody by groups of non-native subjects.6 There is common consent that L1 plays a consistent role in L2 prosody’s acquisition. In this specific view, Chen and Fon7 analyzed the prosodic features of Mandarin speakers’ English production; the reported observations claimed that in L2, both the prosodic grouping patterns and the pitch accents assignment followed

5

For instance, Philippe Boula de Mareüil, Giovanna Marotta and Martine AddaDecker, “Contribution of prosody to the perception of Spanish/Italian accents,” in Bernard Bel and Isabelle Marlien (eds.), Proceedings of the 2nd International Conference on Speech Prosody Nara, Japan, 23-26 March 2004, Nara 2004, 681684, explored the role of segmental and suprasegmental cues in the realization of a Spanish/Italian foreign accent employing a prosodic transplantation tool. Anna Kaglik and Philippe Boula de Mareüil (“Polish-accented French prosody in perception and production: transfer or universal acquisition process?”, in Proceedings of the 5th International Conference on Speech Prosody, Chicago, Illinois 11-14 May 2010, Illinois 2010, Illinois University Press accessed June 2012, http://speechprosody2010.illinois.edu/program.php/papers/100987.pdf) focused on the mastery of prosody by a group of Polish learners in French second language; results suggested that the acquisition of the L2 prosody did not seem to be constrained by the age of the first exposure to the L2. 6 See, among others, Matthias Jilka, “The contribution of intonation to the perception of foreign accent. Identifying intonational deviations by means of F0 generation and resynthesis” (Ph.D. Diss., University of Stuttgart, 2000); Robin M. Queen, “Bilingual intonation patterns: Evidence of language change from TurkishGerman bilingual children”, Language in Society 30 (2001): 55–80; Boula de Mareuil, Marotta and Adda-Decker, “Contribution of prosody”, 681-84; Jürgen Trouvain and Ulrike Gut, eds., Non-Native Prosody. Phonetic description and teaching practice, (Berlin: Mouton de Gruyter, 2007); Mark Swerts and Sabine Zerbian, “Prosodic transfer in Black South African English”, in Proceedings of the 5th International Conference on Speech Prosody Chicago, Illinois 11-14 May 2010, Illinois 2010, Illinois University Press accessed June 2012, http://speechprosody2010.illinois.edu/program.php/papers/100198.pdf. 7 Sally Chen and Janice Fon, “Prosodic features of non-native English production”’ in Wai-Sum Lee and Eric Zee (eds.) Proceedings of the 17th International Congress of Phonetic Sciences. China 17-21 August 2011, Hong Kong 2011: City University of Hong Kong, 456-59.

18

Chapter One

different strategies. Bahler et al.8 considered indeed the role of F0 in listeners’ segmentation of French speech, i.e. the identification of words final boundaries, in native English and native French, divided into two proficiency groups; the findings of a perceptual test, comprehending different synthesized stimuli, proved that only native speakers are able to use F0 rise as a cue to word-final boundary in French. Another line of research analyzed the time effects of L2 exposure on adult learners’ acquisition of prosody and fluency features.9 Finally, a certain number of researches focused on the contribution of prosodic cues in native and non-native subjects in segmenting speech chain into Tus.10 Despite the different aims pursued by these studies, some recurrent traits emerged. First, all studies confirmed that prosody is a strong linguistic feature for speech segmentation, regardless of the language; both native and non-native subjects try to signal acoustically the edges of intonational units, in order to make easier discourse comprehension. However, there is at least one notable difference. During a segmentation task based on listening impression, native speakers can rely on a more varied number of linguistic phenomena, including syntax and semantics; on the contrary 8

Carly L. Bahler, Caitlin E. Coughlin and Annie Tremblay, “Differential contribution of prosodic cues in native and non-native speech segmentation”, in Wai-Sum Lee and Eric Zee (eds.) Proceedings of the 17th International Congress of Phonetic Sciences. China 17-21 August 2011, Hong Kong 2011: City University of Hong Kong, 276-79. 9 Cf. Pavel Trofimovich and Wendy Baker, “Learning second language suprasegmentals: effects of L2 experience on prosody and fluency characteristics of L2 speech”, Studies in second language acquisition 28 (2006): 1-30; Pavel Trofimovich and Wendy Baker, “Learning prosody and fluency characteristics of second language speech: The effect of experience on child learners’ acquisition of five suprasegmentals”, Applied Psycholinguistics 28 (2007): 251–76; Jasone Cenoz, “Pauses and hesitation phenomena in second language production,” ITL: Review of applied linguistics 127-28 (2000): 53-69; Riikka Ullakonoja, “Pausing as an indicator of fluency in the Russian of Finnish learners”, in Plinio A. Barbosa, Sandra Madureira and César Reis (eds.) Proceedings of the 4th International Conference on Speech Prosody, Campinas Brazil, 6-9 May 2008: Campinas, Brazil 2008: Editora RG/CNPq, 339-42. 10 Cf. Irina Nesterenko, “Prosodic annotation in spontaneous Russian: perceptual annotation and automatic classification, speech prosody”, in Rüdiger Hoffmann e Hansiörg Mixdorff (eds.), Proceedings of International 3rd Conference on Speech prosody, Dresden, Germany 2-5 May 2006, TUD PRESS, accessed June 2012, http://aune.lpl.univ-aix.fr/~sprosig/sp2006/contents/papers/PS5-29_110.pdf. Amina Mettouchi et al., “Only prosody? Perception of speech segmentation in Kabyle and Hebrew”, Nouveaux cahiers de linguistique française 28 (2007): 20718; Tyler and Cutler, “Cross- language differences,” 367-376.

Prosodic Features in Native and Non-Native Speech Segmentation

19

non-native subjects rely above all on prosodic cues. At the production level, non-native speakers tend to transfer the prosodic patterns of their mother tongue to the second language, even after long period of L2 exposure. Nevertheless, the amount of prosodic divergence seems to depend, to some extent, on the learner’s level of proficiency.

2. Aim This study deals with the issue of the prosodic phrasing in Italian as second language. Most previous research in the area of L2 prosody’s acquisition has focused on a single dimension. For this reason, the present study had wider aims. Our analysis was performed in both production and perception perspective. The former attempted to explore in detail the phonetic realization of prosodic boundaries, in Italian L1 (It-L1) and Italian L2 (It-L2). The main purpose of this experimental step was to verify whether perceptual results were supported by speech production data. In particular, we investigated the way in which Polish learners, with low proficiency in Italian, use acoustic cues during the production of a prosodic boundary. In other words, the study seeks to ascertain to what extent, native and non-native speakers of Italian differed according to the prosodic boundary marking. The latter analysis intended to test the perceptual role conveyed by acoustic indices in the realization of a prosodic boundary. In this paper we will argue that production and perception investigations are the proper empirical basis to define prosodic phrasing as well as to evaluate the differences emerged in the speech production of unexperienced learners of Italian L2. Listening experiments with Italian natives and with Polish learners were performed to detect intonation boundaries under different experimental conditions. We presumed that the phenomenon under investigation needs special prosodic-acoustic cues and that subjects have a sort of metalinguistic assessment of the prosody of their native language, a prerequisite lacking in non-native learners. The results will be discussed in the following sections.

3. Methodology Two groups of subjects were recruited from a university community: a native Italian-speaking group (It-L1) consisting of three female speakers of Bari Italian aged between 20-23 and a native Polish-speaking one (ItL2) with no previous exposure to the Italian language. These latter were

20

Chapter One

Erasmus students (1 male and 3 females) aged between 21 and 26 years. At the time of testing, all participants were attending university in Bari, a southern Italian town. The Italian proficiency level of the Polish speakers was equal to A2, as performed by certification. Both groups were asked to describe a set of picture postcards, the Polish subjects were also asked to talk freely about their experience of studying in Italy. Digital recordings were made in a suitable silent room by means of a Tascam DR-07 (.wav files, sampling frequency of 22 Khz, 32 bits). The speech material under study consisted of excerpts of spontaneous speech. The corpora were segmented into dialogic turns and then into TUs and labeled with a sequential abbreviation containing the identification of learner’s name, native language and speech typology. The acoustic analysis was accomplished by means of PRAAT. As a whole, we examined 105 TUs for IT-L1 and 120 TUs for It-L2. The following parameters were measured for each TU: (i) syllable length; (ii) articulation and speech rates;11 (iii) the length of pauses and their placement; (iv) temporal duration of stressed and unstressed vowels (considered in both final and non-final position); and (v) pitch range and intonational contour.

4. Acoustic findings 4.1. Speech rate Overall, the verbal production of Italian L2 showed a low degree of textual and prosodic cohesion. It therefore presented a high index of disfluency which was manifested by prolonged use of pauses, interruptions, false starts and repetitions. In order to quantify the degree of disfluency found in the two samples, verbal fluency was considered in a comparative way. With regard to speech rate, L1 and L2 differed. A first indication in this sense comes from the computation in syllables of the temporal extension of TUs. To this purpose, we counted only the syllables phonetically realized. In Italian L2, TUs were shorter. In the same direction, even articulation and speech rates were significantly slower than those of L1 speakers, as shown in Table 1-1.

11

Articulation rate was defined as the number of syllable per seconds, excluding pause time.

Prosodic Features in Native and Non-Native Speech Segmentation

Language

TU

It L1 It L2

6.9 (2.3) 4.8 (2)

Articulation rate 5.1 (1.3) 3.6 (0.9)

21

Speech rate 4.6 (1) 2.7 (0.8)

Table 1-1: Mean values in syllables of TU length, Articulation and Speech rate. Standard Deviation values in brackets. In detail, in Polish subjects TUs included 4.8 syllables while in Italian natives the length of TUs was 6.9. These results clearly indicate that speech rate varied depending on the considered sample (native or nonnative Italian). Still more evident was the difference concerning speech rates. In Polish learners, speech rate was slower, perhaps because of their high rating of pauses. If so, the low speech rate values observed seemed to be directly correlated to pauses, but also to a lower articulation rate. This was only the first macroscopic effect of the high degree of disfluency which characterizes non-native speech production.

4.2. Pauses The number and duration of pauses found in non-native corpora was consistently high. As is known, pausing is also a frequent phenomenon in native language, since it is connected to the physiological mechanisms of breathing, but also to the needs of syntactic and textual planning.12 The pauses are not all alike, for this reason we distinguished silent pauses by filled ones; these latter are manifested through vocalization (i.e. ehm, mhm) or vowel prolongation (i.e. c’è un lago).13 What is interesting to note is that over our entire corpora, all the pauses, regardless of typology, were more numerous and also longer in L2 (Table 1-2). Speech sample It L1 It L2

Silent pauses

Filled pauses

42% 55%

16% 38%

Table 1-2: Percent distribution of pauses by type. 12

Cf. Danielle Duez, “Silent and non-silent pauses in three speech styles”, Language and Speech 25 (1982): 11-28. 13 A silent pause is a moment of silence, conventionally longer than 200 ms, while a filled pause is any occurrence of hesitation.

Chapter One

22

In the Polish learners’ production there was a high percentage of silent pauses, precisely 55%, while in L1 we found 42%. The presence of filled pauses was 16% in L1, but 38% in L2.14 On this point, we noted a considerable variation between the two corpora. In non-native speech, the pauses frequently interrupted the syntactic units, and a pause sometimes interrupted the word. The most conspicuous differences concerned filled pauses, so demonstrating the presence of verbal planning difficulties. This occurs when the learner shows uncertainty about the choice of words, or when the lexical form of a word is not known or temporarily unavailable. Consequently, the greater the number of pauses, the lower the index of textual cohesion. Table 1-3 shows the duration of pauses. Speech sample It L1 It L2

Silent pauses

Filled pauses

500 (270) 580 (241)

616 (280) 613 (261)

Table 1-3: Mean pauses duration (in ms) and Standard Deviation values (in brackets). Coherently, as far the silent pauses were concerned, in L2 there was greater average length; by contrast, the duration of filled pauses was controversial: tendentially silent pauses had a longer duration in L2, but temporal difference was not always appreciable. In It-L2, the length of silent pauses went from 250 ms to 1000 ms; the mean length was 580, and most of the pauses were between 400 and 750 ms. The true disagreement did not just concern the duration of pauses, but rather their statistical impact as well as the functions provided. In Italian L1, many pauses were used for emphasis and stylistic effects,15 while in Italian L2 a great number

14 In native Polish spontaneous dialogues the percent rates of silent pauses is of 49.4% whereas for filled pauses the average result is 22.5% (cf. Maciej KarpiĔski and Ewa J. Nowikow, “Prosodic and gestural features of phrase-internal disfluencies in Polish spontaneous utterances”, in Proceedings of the 5th International Conference on Speech Prosody Chicago Illinois 11-14 May 2010 Illinois 2010, Illinois University Press, accessed June 2012, http://speechprosody2010.illinois.edu/program.php/papers/100152.pdf. 15 These results are consistent with those already found in other research carried out for Italian language. Cf. Massimo Pettorino and Antonella Giannini, “Analisi delle disfluenze e del ritmo di un dialogo romano”, in Italiano parlato. Analisi di

Prosodic Features in Native and Non-Native Speech Segmentation

23

of pauses expressed hesitation and planning problems. Extensive pausing is a typical phenomenon of non-native speech.16 Thus, speech rate decreased and pauses’ occurrence increased. In addition, we noted that in both speech samples, filled pauses may co-occurred with silent ones, but this behaviour was particularly clear in It-L2; in Polish corpus pauses occurred more frequently together with hesitations, self-repairs and repetitions.

4.3. Final lengthening An immediate consequence of the slow speech rate reached in Italian L2 was the increase of vocalic length. Final lengthening of the nuclear stressed syllables was strongly linked with phrase boundaries. In native Italian, a high F0 pitch without phrase-final lengthening did not provide the most optimal prosodic condition for a boundary’s detection;17 however, on this point there were significant cross-linguistic differences. In order to evaluate the presence of temporal divergences, we calculated average vocalic duration by speakers and by speech samples comparing them in a contrastive way. We took into account both stressed and unstressed vowels recurring in final and non-final positions within the utterance. In detail, we distinguished the following four contextual conditions: (1) duration of final stressed vowels (i.e. the last stressed vowel of a TU), F-SV (2) duration of non-final stressed vowels, NF-SV (3) duration of final unstressed vowels, F-UV (4) duration of non-final unstressed vowels, NF-SUV.

un dialogo, eds. Federico Albano Leoni and Rosa Giordano (Napoli: Liguori, 2005), 89-104. 16 Cf. Yoonsook Mo, “Duration and intensity as perceptual cues for naïve listeners’ prominence and boundary perception”, in Plinio A. Barbosa, Sandra Madureira and César Reis (eds.) Proceedings of the 4th International Conference on Speech Prosody, Campinas Brazil, 6-9 May 2008: Campinas, Brazil: 2008, Editora RG/CNPq, 739-42. 17 See for Italian language: Patrizia Sorianello, “Per una definizione fonetica dei confini prosodici”, in Massimo Pettorino, Antonella Giannini and Renata Savy, (eds.), The International Congress of La comunicazione parlata, Naples, February 2006 (Naples: Liguori, 2008), 310-30.

Chapter One

24

The temporal findings of both stressed and unstressed vowels were compared with those obtained for non-native corpus, as clear from data summarized in Table 1-4. Duration tendencies seem to be conditioned by the native or non-native speech condition. Our data showed that in L2 the final unstressed vowels had the longest duration. This suggested that the duration of the unstressed vowels is a significant predictor factor capable of detecting the presence of a prosodic boundary. Speech Sample It L1 It L2 t-test

NF-SV

F-SV

NF-UV

F-UV

75 (25) 126 (39) 0.001

170 (46) 180 (30) 0.177

60 (16) 80 (17) 0.054

115 (32) 184 (44) 0.001

Table 1-4: Mean temporal values (in ms) and Standard Deviation (in brackets) of stressed and unstressed vowels in final and non-final position for Italian L1 and L2, t-test values. In L1, stressed vowels were always longer with respect to the unstressed ones occurring in the same context. As expected, when the vowel, indifferently stressed or unstressed, was in prepausal position, its length was appreciably high. Nevertheless, in Italian L2 the results showed a reversal trend. In fact, in the Polish learners the more important lengthening concerned the final unstressed vowels; diversely the nuclear stressed vowels underwent a minor temporal increase.18 Additionally, in L2, F-SV and F-UV had a comparable length (i.e. 180 ms vs. 184 ms). For a more detailed analysis, it is worthwhile to note that the temporal increment does not spread over all the syllable nuclei, probably for the presence of a reduced speech rate. In IT-L2, the greatest amount of lengthening, with respect to L1 sample, was achieved by NF-SV and FUV. Regarding the NF-SV, it is reasonable to assume that learners had paid much attention to the realization and placement of lexical accent. This 18

It is important to observe that, from a rhythmic point of view, Polish displayed intermediate properties between stress-timed and syllable-timed languages. Acoustic measurements and perceptual experiments seem to suggest for Polish the postulation of a new rhythm class. Cf. Franck Ramus, Emmanuel Dupoux and Jacques Mehler, “The psychological reality of rhythm classes: perceptual studies”, in Maria Josep Solé, Daniel Recasens and Joaquín Romero (eds.), Proceedings of the 15th International Congress of Phonetic Sciences, Barcelona, Spain 3-9 August 2003, Barcelona 2003, Universitat Autonòma de Barcelona, 337-42.

Prosodic Features in Native and Non-Native Speech Segmentation

25

behaviour was not unexpected: the two languages considered (Italian and Polish) differed as far the stress was concerned. Polish is a fixed-stress language, since it generally has a penultimate stress. In contrast, Italian is a free-stress language: for this reason, primary stress may be used contrastively. Besides this stress-conditioning factor, another variable has to be considered. In It-L2 the unstressed final vowels also reached high temporal values. We think that in Italian L2 the difficulties encountered by learners during the communicative situation and the overall disfluency level were clearly correlated. Consequently, the lengthening of final unstressed vowels may be determined not only by the prepausal position, but also very often by recurrent hesitations and, to a wider extent, by problems planning. In most cases, we found that lengthenings themselves constitute a disfluency.

4.4. Intonation contours With respect to this, our attention focused on pitch variations produced in the vicinity of an intonational boundary. Comparing L1 and L2 samples, a diverging situation appeared. In native Italian, the range of final F0 intonation patterns was varied and closely linked to the terminal or nonterminal nature of Tus.19 The unmarked conclusive pattern, typically that of an assertive utterance, had almost always realized by a falling movement (in ToBI transcription !H+L* L-L%), in a smaller number of cases we found a mid-level F0 pattern not reaching the baseline which continues until the end of the TU. A rising movement characterized the continuation TUs. These data confirmed what previously observed for Bari Italian.20 19

The non-terminal contour corresponds to a continuation rise, a typical gradual rise signaling that the utterance is not ended and ‘more is to come’. Thus, it expresses uncertainty and non-finality. On the contrary, the terminal F0 contour means that a TU is over, it conveys a general meaning of conclusion, closeness and finality. For more investigations on the fall/rise F0 patterns distinction, see Cruttenden, “Falls and rises”, 77-91. 20 Cf. Michelina Savino, “Non-finality and Pre-finality in Bari Italian Intonation: a Preliminary Account”, in Paul Dalsgaard, Børge Lindberg (eds.), Eurospeech 2001 Scandinavia: 7th European Conference on Speech Communications and Technology, 3-7 September 2001, Aalborg, Denmark 2001, Kommunik Grafiske Løsninger A/S, 939-42; Michelina Savino, “Intonational Cues to Discourse Structure in a Variety of Italian”, in Regional Variation in Intonation, eds. Peter Gilles and Jörg Peters (Tübingen: Niemeyer, 2004), 145-59; Michelina Savino et al., “Intonational cues to discourse structure in Bari and Pisa Italian: perceptual evidence”, in Rüdiger Hoffmann and Hansiörg Mixdorff (eds.), Proceedings of the

26

Chapter One

The speech of Polish learners showed an altered distribution of melodic configurations which immediately transmitted a listening impression of ‘foreign accent’. The intonational inventory was certainly less rich than in It-L1. The falling contours were much reduced in number, only 15%, while the percent values of rising and level configurations increased. A falling pattern was unusual also for conclusive TUs. In Fig. 1-1 we represented a spontaneous utterance produced by a Polish subject in which the same contour was employed for both a continuative TU (i.e. ma la domenica) and a conclusive one (i.e. noi andiamo a Roma).

Fig. 1-1: F0 pattern of a spontaneous utterance produced by a female native Polish speaker; : silent pause.

In 29% of all cases, the intonational boundary was realized through a sort of plateau, i.e. a pattern ending with a sustained pitch placed around the mid point in the speaker’s range. The pitch remained at a mid level realizing an unfinished TU, a stretch of speech intonationally open. Typically, this flat movement occurred when the TU contained prolonged or corrections of lexical elements. Emblematically, in a certain sense, the verbal uncertainty has been reflected through the prosodic choice of a less definite melodic contour, neither clearly rising, neither clearly falling. There was an interesting trend in some of the L2 data analyzed so far. Our research documented a growing usage of high-rising terminal patterns. Thus, in the Polish learners’ Italian, more than half of terminal 3rd International Conference on Speech Prosody, Dresden, Germany, 2-5 May 2006, Dresden 2006, TUD PRESS, 114-17; Martine Grice et al., “Strategies for intonation labelling across varieties of Italian”, in Prosodic Typology: the Phonology of Intonation and Phrasing, ed. Sun-Ah Jun (New York: Oxford University Press, 2005), 362-89.

Prosodic Features in Native and Non-Native Speech Segmentation

27

TUs ended with a rising trend (56%). This final melodic movement associated with a high boundary tone (H%) transmitted a well-defined continuation meaning. However, the pattern was different, in form and distribution, if compared to that found out in native Italian, since the rise covers only the time domain of the final unstressed syllable. There have been no cases in which the rise began on the nuclear vowel; this last was therefore implemented on a very low F0 level. On average, the amount of rising movement was significant, the frequential difference calculated in ST between the value of F-SV and F-UV being equal to 6.3 (SD 2.7). A comparison between the two diagrams shown in Fig. 1-2 led to the following observations. In It-L2 the last stressed syllable of the TU was low (L*) followed by a rise on the subsequent unstressed one (H%), while in the southern variety of Italian the nuclear syllable was strongly falling (H+L*) or sometimes lightly falling (L*). In both speech samples, the boundary tone was high (H%), but the tonal excursion differed, namely 6 ST in L2, but 4 ST in L1.

Fig. 1-2: Tonal configuration of continuation rise in Italian L1 (left) and Italian L2 (right).

In native Italian there was a good degree of variability in the domain of terminal and non-terminal F0 contours. This was not surprising: as expected, native speakers have a thorough knowledge of their mother tongue. Thus, they succeeded in expressing a great deal of pragmatic functions, also varying lightly their F0 contours. Native speakers were able to choose the pitch patterns according to the context and to the information function they wanted to communicate. On the contrary, in ItL2, intonational configurations were rather monotonous. The most reliable F0 changes were realized on the right edge of the intonation phrase, that is, in the correspondence of a boundary tone. Sometimes, the F0 variations did not occur, hence, as already reported above, the melodic movement

28

Chapter One

was expressed by a flat and sustained mid-level pitch which stretched over a long portion of the utterance (see Fig. 1-1). Usually, learners tended to superpose their native intonation patterns on the second language; such prosodic interference, of course, contributed to increase the general degree of foreign accent. It should be remarked that prosodic transfer is a particularly resistant phenomenon also recurrent in high second language proficiency.21 Nevertheless, our data lent itself to other considerations. The statistical incidence of rising patterns, even in those sites where it should be expected a conclusive contour, seems to signal a specific pragmatic function.22 Learners continuously asked the native speakers for confirmation thus showing a subordinate attitude towards them. This proves they are aware that their proficiency of Italian language is still inadequate. The function of the rising contour seems to be a way for delimiting the TUs which have not been completed. In this way they are signalling that there is still more to come. The high pitch reflects, in a certain sense, the planning difficulty encountered by learners in continuing the conversation. A final rising contour transmits a sense of opening and incompleteness; and that in all languages fall and rise contours establish a contrastive melodic meaning.23 Thus, in most languages the paralinguistic meanings of falls and rises, i.e. the choice of high and low F0 contours, undergo a process of grammaticalization. Moreover, this rising intonation pattern was used predominantly by Polish subjects to find out from the native speakers whether they had been 21

Cf. Jilka, “The contribution of intonation”, 1-225; M. Queen, “Bilingual intonation patterns,” 55–80; Kaglik and Boula de Mareüil, “Polish-accented French prosody”. 22 The studies on Polish intonation have not a long tradition; researches on spontaneous speech are also limited. See Wiktor Jassem, “Computer-based classification of basic Polish intonations”, in Proceedings of the 11th International Congress of Phonetic Sciences, Tallin, Estonia 1-7 August 1987, USSR, 253-56; Katarzyna Francuzik, Maciej KarpiĔski and Janusz Klesta, “A preliminary study of the Polish intonational phrase, nuclear melody and pauses in Polish semispontaneous narration”, in Bernard Bel and Isabelle Marlien (eds.), Proceedings of the 1st International Conference on Speech Prosody, Aix-en-Provence, Avril 11-13 2002, Aix-en-Provence 2002, Laboratoire Parole et Langage, 303-06; Maciej KarpiĔski, “The corpus of Polish intonational database. Technical specification”, in Investigationes Linguisticae VIII (2002): 7-8; Esther Grabe and Maciej KarpiĔski, “Universal and language-specific aspects of intonation in English and Polish”, in Maria Josep Solé, Daniel Recasens and Joaquín Romero (eds.), Proceedings of the 15th International Congress of Phonetic Sciences, Barcelona, Spain 3-9 August 2003, Barcelona 2003, Universitat Autonòma de Barcelona, 1-4; KarpiĔski and Nowikow, “Prosodic and gestural features”. 23 Cf. Cruttenden, “Falls and rises,” 77-91.

Prosodic Features in Native and Non-Native Speech Segmentation

29

understood, and whether they had used the correct forms in speaking; the learners are continually asking for agreement about the correctness and adequacy of the realized verbal message Polish learners thus used an F0 rising contour to signal a request for help.

5. Perceptual test A prosodic boundary is the result of the complex acoustic aggregation of more phonetic features. Each of them provides a different perceptual effect in the spoken language. There were essentially three underlying hypotheses that guided the present research: (i) which is the perceptual role carried out by each prosodic cue in the boundary’s realization? (ii) Are such prosodic features language-specific? (iii) Which are the perceptual consequences derived by the artificial neutralization of some prosodic indexes? In order to verify the auditive impression conveyed by two different speech samples (namely, native and non-native) during the prosodic segmentation of spoken flow, a perceptual test was accomplished.

5.1. Stimuli For the perception experiment, 16 speech stimuli of variable length were selected from our corpora: eight corresponding to native Italian and eight corresponding to non-native Italian. We adopted two different experimental strategies: the first eight excerpts (4 concerning It-L1 and 4 concerning It-L2) were presented in their natural sounding version, while the remaining ones, different from the former in order to avoid a ‘memory effect’, were manipulated according to the following prosodic conditions: (i) No Prosody: the stimuli were presented with flattened F0 coinciding with the mean value of F0 previously computed for each utterance. As a consequence, all intonational variations were neutralized and the F0 curve was perfectly monotonous. (ii) No Pausing: both silent and filled pauses were removed from the original audio signals, while F0 contours were left unchanged. (iii) No Prosody + No Pausing: the neutralization of F0 contours and the removal of pauses were performed.

30

Chapter One

5.2. Methodology A group of 10 Bari listeners with no previous experience of prosody took part in the experiment; the informants did not know the purpose of the perception test. All the participants were tested together in a quiet room using a personal computer and software for the presentation of stimuli. The re-synthesized stimuli were obtained by means of the algorithm PSOLA (Pitch-Synchronous Overlap-Add). The analysis was based on 32 speech stimuli, respectively 16 for L1 and 16 for L2 (8 were natural versions, whereas the other 24 were artificial versions). We also employed 8 distracter sentences. The recorded stimuli selected for the perception test were presented via computer in a random order; and each stimulus was followed by five seconds of silence. All speech stimuli, both natural and synthesized, were orthographically transcribed in a random order on single sheets, the printed texts being presented without punctuation marks. Listeners were asked to listen to the audio portions and to mark on the printed pages the position of intonational boundaries. No explicit definition of the boundary was proposed, subjects were asked to put a separation index at the places where they heard a sort of discontinuity among words. The judgements were made solely on the perceptual basis, in real time as they listened. The detection rates were compared with the boundary markers previously determined by the phonetic analysis. The perceptual analysis was instead carried out on 320 responses (32 stimuli x 10 listeners’ judgements). The findings of the perception experiment mirrored the degree of importance of the prosodic cues under investigation.

5.3. Results The perception task of having to mark the boundary between consecutive TUs was more duifficult than it appeared. Boundary identification was given by the percentage rates of subjects agreeing on a given break. Nevertheless, listeners used different perceptual strategies according to native or non-native typology of the listened stimuli. Generally, listeners’ perceptions were not only influenced by prosody, but by overall fluency degree. First inspection revealed that experimental data of Italian L1 and Italian L2 were comparable, since both presented similar trends. The phonetic realization of a boundary tone did not seem to be language-specific. Moreover, the two speech samples differed in their percent rates of boundary’s detection. Results demonstrated the presence of substantial differences between the two corpora. Tables 5-8 gave the results of detecting intonational

Prosodic Features in Native and Non-Native Speech Segmentation

31

boundaries in various experimental conditions. First of all, let’s consider Italian L1. The original audio signals were correctly identified by listeners with a percentage of 88% for conclusive TUs, but only of 55% for continuative ones. This confirmed what had already been found in other studies,24 that the TUs conveyed a conclusive meaning, usually identified by a falling pattern, showed major perception strength. In the first experimental condition (No prosody), the right identification of an intonational break fell drastically to the 34%, confirming, as it is widely known, that intonation is a very reliable feature for the prosodic phrasing of connected speech. In the second condition (No pausing), the removal of pauses yielded a better result, namely 60%. As expected, when considering the third condition (No prosody + No pausing), there was a shift in response; both manipulation of pauses and F0, got worse, the relative percentage was in this case of 28%. In this specific experimental condition, listeners failed to carry out the annotation of prosodic phrasing, see what listened in Tables 1-5 and 1-6. Thus, as revealed by perceptual responses, at least ideally, the perceptually preferred condition for the boundary tone’s detection in native Italian is that marked by an intonational contour change. Italian L1 Original audio signals Continuation TUs 55% Conclusive TUs 88% Table 1-5: Percent rates of detecting intonational boundaries for L1 natural sounding signals.

24

Cf. Cruttenden, “Falls and rises”, 77-91; Mark Swerts, René Collier and Jacques M.B. Terken, “Prosodic predictors of discourse finality in spontaneous monologue”, Speech Communication 15 (1994): 79-90; Mark Swerts, Don G. Bouwhuis and René Collier. “Melodic cues to the perceived ‘finality’ of utterances”, Journal of Acoustical Society of America 96 (1994): 2064-75; Aoju Chen, “Languagespecificity in the perception of continuation intonation”, in Tones and tunes II: phonetics and behavioural studies in word and sentence prosody, eds. Charles Gussenhoven and Tomas Riad (Berlin: Mouton de Gruyter, 2007), 107-42.

32

Chapter One

Italian L1 Synthesized stimuli No Prosody No Pausing No Prosody + No Pausing

34% 60% 28%

Table 1-6: Percent rates of detecting intonational boundaries for L1 artificial speech signals. Italian L2 Original audio signals Continuation 84% TUs Conclusive TUs 95% Table 1-7: Percent rates of detecting intonational boundaries for L2 natural sounding signals. Italian L2 Synthesized stimuli No Prosody No Pausing No Prosody + No Pausing

82% 51% 36%

Table 1-8: Percent rates of detecting intonational boundaries for L2 artificial speech signals. In Polish speech production we observed a strong alteration of the prosodic organization. At first glance, experimental findings seemed surprising. With regard to L2, the percentage values of prosodic boundary identification were higher with respect to those obtained for native informants. Therefore, when subjects listened to non-native Italian, the speech segmentation task appeared easier. This does not mean that Italian listeners adopted different prosodic segmentation strategies, according to the nature of speech stimuli (i.e. native or non-native), but rather that Italian language realized by the Polish learners was very different by that of the natives. The percent incidence and distribution of the acoustic indexes considered were in fact diverse. Tables 1-7 and 1-8 revealed

Prosodic Features in Native and Non-Native Speech Segmentation

33

striking results concerning L2 speech segmentation in different experimental conditions. During the identification of It-L2 natural signals, the Bari listeners achieved higher values, particularly as regards continuative TUs; the best detection rates of the non-terminal TUs can probably be ascribed to the presence of an atypical continuation rise (as shown in Fig. 1-1). The presence of a rapid F0 rise represented a valid listening predictor of a prosodic boundary, in that it has successfully predicted a boundary tone. With regard to synthesized stimuli, the results achieved showed that Italian listeners performed better with the no prosody condition. The manipulation of the F0 parameter alone did not significantly change the auditory judgements. Listeners correctly identified 82% of the boundaries. Therefore, fundamental frequency was not a decisive feature, although it has played a certain role in the detection of TUs. Thus, unexpectedly, its influence appeared marginal. In other words, also taking into account what happened in the second experimental condition, it was clear that intonation contributed above all when information pauses were absent. Looking more precisely at the data gained in the no pausing condition, we saw that the effect produced by the removal of pauses was relevant. The percent values of well-identified prosodic boundaries dropped to 51%. The artificial suppression of pauses improved the fluency of the verbal message. Thus, the utterances became faster and more fluent. In any case, in a fast articulation rate, prosodic boundaries are always fewer. The absence of pauses was the most prominent factor in determining low rates, since it confused the listeners’ judgements. Consequently, their own hearing performance significantly worsened. This means that in non-native speech, the silence information helped listeners to detect boundaries, improving overall precision. Acoustic data indicate that pauses, and disfluency in general, are strong predictors of non-native speech, In fact, by manipulating pauses, listeners’ perplexity increased. In such a way, the high number of speech interruptions and slowdowns created a less fluent impression of the spoken chain. Finally, in the no prosody + no pausing condition, the manipulation of two parameters (F0 + pauses) acted in a synergic way decreasing the percent values of boundaries’ detection. In It-L2, pitch configurations and speech disfluencies interacted with the identification of prosodic markers. We found evidence that pausing, hesitations, and self-repairs are reliable acoustic cues able to affect the auditory judgements of Italian subjects, since they contributed to delimiting the prosodic region of a TU. Learners with low expertise of a second language, like our Polish participsants,

34

Chapter One

were inclined to accumulate more disfluency phenomena, thus leading to a kind of ‘interference effect’ when compared with the natives.25

6. Discussion and conclusion This study suggests that native and non-native Italian corpora, besides some prosodic convergences, show several divergences, from the twofold point of view of production and perception. The results allow us to draw conclusions at different levels. First of all, the existence of a significant difference between native and non-native speech condition and prosodic parsing was confirmed. Having established that there is a positive correlation between language proficiency and the overall level prosodic segmentation, it remains to be explained to what extent the two variables are linked. Learners do structure the speech chain through particular intonational shapes and by pauses organization. What occurs in non-native speech is a kind of prosodic disorganization which is realized through short TUs, long moments of silence, and the insertion of non-verbal lexical items. There are several implications arising from our experimental outcomes. To this end, it should be remarked that fluency was not only given by the number of the words produced within an utterance or a TU, but rather by the joint activation of more prosodic parameters. When the verbal competence of a second language is still lacking, learners tend to transfer in L2 the intonational and pragmatic functions of their native language. In our speech sample of Italian L2, the acquisition of prosody occurred late and learners’ adequacy is not complete. The acquisition of a native-like speech in fact never does occur in the early years of exposure to a second language. However, prosodic transfer is not the only explanation: we came to believe that proficiency is one of the determining variables for prosodic boundaries’ detection. In Polish learners we observed a notable speech over-segmentation, determining a prosodic discontinuity effect. Textual organization of the verbal message has not yet been mastered by Polish learners; the reduced size of TUs, often limited to a single word, reveals serious problems impeding the overall planning of the utterance: the simultaneous control of both semantic content and form for these learners is a very exacting task. If so, it may be

25

Unfortunately, in this research we had no evidence for the temporal lengthenings consequences, because all stimuli preserved their original durational patterns.

Prosodic Features in Native and Non-Native Speech Segmentation

35

useful in future research to examine the phenomenon of prosodic phrasing using highly experienced learners.26 Interestingly, production data in L1 are only partially coincident with perceptual responses, while in L2 there is a better coincidence. On the other hand, on perceptual outlook, we observe that listeners are able to discriminate TUs from the two languages on the basis of prosodic cues. Our study proves that perceptual experiment represents a valid methodological device revealing all the perceptible differences conveyed by F0 variations and pauses placement. In both corpora, the two acoustic parameters used to create the artificial conditions reinforced each other. However, melody and pausing in Polish speakers do not appear to induce the same perceptual consequences. There is good reason to postulate that non-native Italian has a different prosodic organization. Listeners integrated information from multiple acoustic cues in the detection the presence of a prosodic boundary. In native speakers, the predictability of a boundary marker is influenced by a lot of variables including also syntactic, semantic and pragmatic meanings. Diversely, in non-native subjects, the predictability is rather given by the density of disfluency phenomena and by the presence of atypical prosodic features. These findings call for future research to better understand the strength of acoustic factors in influencing the perception of L2 prosody. Obviously, much work remains to be done. It would be desirable to extend the investigation to other learners than the Polish ones—i.e. with different mother tongues—as well as to learners with different proficiency levels of Italian.

26

Previous studies demonstrated that fluency improves as the L2 experience increases and that learners’s fluency develops during their exposure to a foreign language, cf. Ullakonoja, “Pausing as an indicator of fluency”, 339-42.

CHAPTER TWO RYTHMIC-PROSODIC ASPECTS IN THE ACQUISITION OF ITALIAN AS A SECOND LANGUAGE: A TEACHING EXPERIENCE WITH ADULT FOREIGN LEARNERS IN CATANIA MILENA ROMANO UNIVERSITY OF CATANIA, ITALY

1. Introduction Starting from the convergence of the studies on theoretical models of average-use Italian1 and neostandard Italian2 applied to the experience of language teaching, we will evaluate here the relevance of the rhythmic and prosodic aspects in teaching Italian as a second language in diversified contexts. Two teaching experiences at different levels, namely courses for university students (2008-2010) and language lessons at an adult education centre sponsored by the Municipality of Catania, Sicily (2009-2010), will be compared. Data were collected through classroom recordings and interviews. The two groups were composed of: a) Erasmus students studying Italian for purposes of cultural enrichment (STUD); b) Adult immigrants studying Italian for work reasons (IMM). 1 Francesco Sabatini, “L’’italiano dell’uso medio’: una realtà tra le varietà linguistiche italiane”, in Gesprochenes Italienisch in Geschichte und Gegenwart, ed. Gunther Holtus and Edgar Radtke (Tübingen: Narr, 1985), 154-83. 2 Gaetano Berruto, Sociolinguistica dell’italiano contemporaneo (Roma: La Nuova Italia Scientifica, 1987).

Chapter Two

38

STUD (Reference Group)

IMM (Control Group)

Courses of History of the Italian Language at the Faculty of Arts and Philosophy, University of Catania School “Federico II” (Private School of Italian Language and Culture in Catania) Learners: 15 students aged between 18 and 25

Adult Education Centre “Alessandro Manzoni” sponsored by the Municipality of Catania

STUD1: university students STUD2: private school students = studying Italian for cultural purposes Origin: Croatia (2), France (4), Germany (3), Spain (4), USA (2)

Learners: 25 adult foreigners with up to pre-high-school education (licenza media) IMM1: in Italy for a short time (16 months) IMM2: in Italy for a long time (6 months -2 years) = studying Italian for work reasons Origin: Bangladesh (4), Bulgaria (1), Burkina Faso (1), China (7), Kosovo (1), Mauritius (3), Morocco (1), Nigeria (1), Palestina (1), Romania (2), Senegal (1), Slovenia (1), Tunisia (1).

Table 2-1: Groups.

1.1. Corpus description The STUD group (reference group) was composed of students of both European and non-European origin (USA), aged between 18 - 25 years old. The European students came to Italy following their participation in the “Erasmus” programme, while the American students were in Italy to attend Italian language courses at public universities or private schools. The observation of linguistic data collected from the STUD group allowed for the consideration of teaching times and methods in contexts where learning L2 Italian is not a necessity, but an enrichment of one’s educational profile. The other context is the IMM group (control group). This allowed for the consideration of linguistic education intended for

Rhythmic-Prosodic Aspects in the Acquisition of Italian

39

adult foreign learners for whom acquisition of L2 Italian is a vital need to interact in the territory they opted to live in, even though only on a temporary basis in some cases. The class group was composed of 25 students and was rather heterogeneous. In most cases, the members of this group felt the need to learn Italian for reasons connected to work, both those who were looking for a first job and those willing to improve their employment status. We compared the two analyzed samples using their approach to Italian (relationship and interaction) as a rationale. Making use of a metaphor deriving from linguistic terminology, we could say that two concepts took shape: “learning as necessity” and “learning as luxury”, namely a need for the students of the IMM group and a luxury for the students of the STUD group.

1.2. The analytical framework selected We know how much the variety of the approaches has characterised the progress of these studies. Of the three levels to be taken into account when discussing intonation—the phonic-acoustic level, the perceptive level, and the linguistic level—we selected perceptive analysis because it allows some identification criteria of the prosodic phenomena to be rendered explicit.3 It is also useful to recall that the term “perception” does not refer to the hearing reaction of the speakers, but to the interpretation they give to linguistic stimuli recognised as such.4 In this sense, we observed the learning of prosodic structures of Italian as a second language focussing our attention on: 1) The role of intonation as a semantic-syntactic separation tool in the process of understanding the statements; 2) The process of acquisition of the rhythmic-prosodic traits of Italian upon which the acceptability of the linguistic act and, consequently, successful communication depend;5 3) The levels of awareness in the utterance of rhythmic-prosodic traits in L2 Italian by the foreign learner.

3

Miriam Voghera, Sintassi e intonazione dell’italiano parlato (Bologna: Il Mulino, 1992), 91. 4 Pier Marco Bertinetto, Strutture prosodiche dell’italiano. Accento, quantità, sillaba, giuntura, fondamenti metrici (Firenze: Accademia della Crusca, 1981), 55. 5 Emanuela Cresti, Corpus di italiano parlato (Firenze: Accademia della Crusca, 2000).

Chapter Two

40

In particular, we tried to investigate the learners’ explicit awareness of the tonal features that allow us to distinguish and classify an intonational contour as appropriate to express an intention or will in a given communication situation.6 With these objectives in mind, data were collected through classroom observation, recordings of spontaneous speech without the students being aware of the recording, and through a “linguistic” interview structured in three parts.

1.3. Structure of the linguistic interview The first part of the interview consisted in collecting the personal and socio-linguistic data of the foreign subjects to define their socio-cultural identities as well as their relationship with Italian as a second language. This part included questions such as: Which is your first language (mother tongue)? How long have you been studying Italian? Where did you study Italian (in your Country of origin, in Italy, at university courses, or other…)? Is there anyone who speaks Italian in your family? The second part tested the learners’ intonational and rhythmic-prosodic meta-skills (phonetic awareness) through questions on subjective perception of real situational speech during their stay in Italy. This was obtained through questions like these: Do you always understand when you are asked a question? Do you understand the meaning of words or the tone with which the question is asked? Do you understand when you are being invited (or urged) to do something? Do you understand if someone is amazed by anything? Do you understand if someone who is speaking is angry? This part of the questionnaire was prepared in order to investigate the relationship between intonation and semantics, considering two different interpretations of this relationship: a)

Intonational variations influenced by the syntactic organisation of the sentence;͹ b) Intonational manifestation of the “emotional” attitudes of the speakers (doubt, anger, etc).ͺ 6

Amedeo De Dominicis, Intonazione e contesto (Torino: Edizioni dell’Orso, 1992), IX. 7 Emanuela Cresti and Massimo Moneglia, eds., C-ORAL-ROM Integrated Reference Corpora for Spoken Romance Languages Studies, University of Florence, in Corpus Linguistics, 15 (Amsterdam: John Benjamin Publishing Company, 2005). See also Patrizia Sorianello, Prosodia (Roma: Carocci, 2006).

Rhythmic-Prosodic Aspects in the Acquisition of Italian

41

The third part of the questionnaire was built on perceptive experimentation conducted using a collection of speech broadcast on television. In this way9 we assessed the didactic function of television texts taken from different types of programmes, and used them as indicators of socio-communication perception and potential models for the learning of situational language. During the linguistic interview, we listened to a number of television texts selected from the three major macro-kinds (news, entertainment, educational), and from a hybrid macro-kind that provided examples of simulated spoken language (in marked diatopy: the American soap-opera Beautiful and the Italian Un posto al sole). We preferred to rely on hearing for our prosodic-intonational analysis and worked at a perceptive level (a less modern, yet more consolidated approach, perhaps).10 We were able to record the effective feedback of the acoustic element in the person learning Italian as a second language, obviously bearing in mind the high risk of subjective evaluation.

1.4. The socio-linguistic framework The socio-linguistic framework adopted, both for the observation of the learning/teaching processes and for data presentation, enabled us to examine in depth acquisition and levels of suprasegmental skills in Italian as a second language using the axes of diatopic, diastratal, diamesical and diaphasical variations. In line with the Common European Framework of Reference for Languages, the opening question was as follows: if it is true that different variation levels cross and shape the Italian linguistic repertoire in terms of phonetic, morpho-syntactic, lexical and syntactic traits,11 how do such “variational features” affect the rhythmic-prosodic traits during acquisition of linguistic skills in L2 Italian?

8

De Dominicis, Intonazione e contesto, IX. Based on Pierangela Diadori, L’italiano televisivo (Siena: Bonacci, 1994); Antonella Giannini and Massimo Pettorino, “Il parlato dei mass media: analisi multilingue del parlato dei telegiornali”, in La Comunicazione Parlata 3, eds. Massimo Pettorino et al. (Napoli: Università degli studi di Napoli l’Orientale, 2010), II. 10 Pier Marco Bertinetto and Emanuela Magno Caldognetto, “Ritmo e intonazione”, in Introduzione all’italiano contemporaneo. Le strutture, ed. Alberto Sobrero (Roma/Bari: Laterza & Figli, 1993), 141-92. 11 Gaetano Berruto, “Varietà diamesiche, diafasiche, diastratiche”, in Introduzione all’italiano contemporaneo. La variazione e gli usi, ed. Alberto Sobrero (Roma: Laterza, 1993), 37- 92. 9

Chapter Two

42

Socio-linguistic variations 1. Diatopic variation 2. Diastratal variation 3. Diamesical variation 4. Diaphasical variation

Suprasegmental skills Prosodic intonational traits of L2 Italian observed during the processes of language teaching/learning.

Table 2-2: Variations.

2. The diatopic component Starting from the first variation parameter (variation in space), we can say that the diatopic axis in particular proved to be a privileged observation parameter in a dialect-marked context as Sicily, in order to measure the incidence of the rhythmic-melodic component in learning the metrics of information contents within a statement. It should be specified that the term “diatopic” acquires a double meaning, since it refers both to the superstratum interlinguistic component (the influences of regional or dialectal traits in the acquisition of Italian), and to the substratum interlinguistic component (mother tongues of the foreign students). Thus, in an interlinguistic framework, we have foreign languages belonging to the learners’ native skills on the one hand, and the learners seen as bearers of linguistic space on the other.

2.1. The Internal Diatopic Component As for the internal diatopic component, intrinsic to Italian linguistic skills, the picture was clearly rather complex from the beginning. It has already been widely acknowledged that, given the peculiar Italian situation, our linguistic repertoire is not consistent throughout Italy, but changes from region to region.12 The intonation of Italian statements is strongly influenced by regional or diatopic variations. It is patently clear to Italians and to foreigners visiting the country that people in the different regions and areas diversify the “intonational contour of their speech” by applying the trend of their typical local dialect to the regional variant of

12

All this creates a problem relating to the reference language-model. Without dwelling here on the theoretical definitions of the concepts of diglossy, bilinguism, dilalia, we will keep them in mind as methodological reference parameters, focusing our attention on the consequences that the socio-linguistic complexity of Italian causes in teaching it as a second language.

Rhythmic-Prosodic Aspects in the Acquisition of Italian

43

Italian.13 In both the STUD group and the IMM group the dialectal or regional component exerted a very strong, sometimes “almost exclusive”, action in their understanding and acquiring the rhythmic-prosodic traits of L2.

2.1.1. The STUD Group and the internal diatopic component: Real cases The questionnaire administered enabled us to more closely identify which were the diatopically marked rhythmic-intonational traits perceived by the foreign learners of the STUD group.14 The question, In your opinion, how do Sicilians (Catania citizens) speak? received different answers, but a high percentage (75%) of STUD (reference group) opted for a statement of this kind: “Sicilians ‘break’ words and leave the last syllable suspended.” Such an answer, even though formulated at a merely perceptive level, highlights rhythmic-intonational traits evaluated as different from the target language being learnt (standard Italian). These observations are also corroborated by other statements given by the subjects. The question, “If you ask someone in the street for a piece of information, do you understand from the answerer’s way of speaking (not from words) whether that persons speaks Italian or dialect?”, received answers such from the interviewees as: MD: Yes, I recognise persons speaking dialect because they mumble and do not utter the last letters; for example, they don’t say “cammino” (I walk), but [ka’m:ingts] [imitating the person], something changes, I don’t know, just like they were slurring the last vowels.

We noticed that, in this case, the phonetic traits peculiar to the regional variant (such as minimum specification of vowels at word end) affect the reception of entire statements. We find here minimum segmental specification, which can result in phenomena like sandhi deriving from coarticulation, elision of phones or syllables, and weakening of consonantal articulations. These facts are peculiar to a particularly shabby and diaphasically low speech and to some diatopic variants such those of

13

Ilaria Bonomi, “La lingua dei quotidiani” in La lingua italiana e i mass media, eds. Ilaria Bonomi et al. (Roma: Carocci, 2003), 131. 14 The initials first (or after) reported sentences (MS, MD, MT, AG, etc.) belong to the people interviewed whose full names are not shown for reasons of privacy.

44

Chapter Two

Southern Italy as Sicilian language.15 These phenomena of low signal specification can affect also morphologically significant segments,16 as registered by some foreign learners: MT: You Sicilians replace the last vowels with a “U”. “Yes, but you don’t say ‘uuuu’, it sounds as you have just slurred it. Now that we know it, we can understand you, but it is always difficult.

In this case, you can observe that the convergence of all the velar vowels at word end into a –u, a phonetic trait typical of Sicily17 is perceived as a prosodic trait (“it sounds as you have slurred the end of the word”). In this regard, notice that a sound can be taken as more or less prolonged without it necessarily having the property in question.

2.1.2. Dialect, idiolect and articulation rate As far as articulation rate is concerned, it can be seen that 45% of subjects perceive the statements uttered in dialect as being faster than those in Italian (Sicilian and Italian articulation rate is the same for 35% of interviewees; Sicilian articulation rate is slower than Italian for 20% of interviewees). In some cases, the double variant (standard Italian and local Italian) available to a speaker is characterised by different rates. Indeed, through the questionnaire we found that, according to foreign learners’ hearing, the same Italian speaker speaks faster when speaking dialect rather than Italian. One student even ventured a possible explanation: “Maybe he/she should think more when speaking Italian” (MD). This perceptive information is confirmed by Romano and Interlandi regarding inter-correlations resulting from cross-checking the utterances of

15 See Miriam Voghera, “Teorie linguistiche e dati di parlato”, in Dati empirici e teorie linguistiche. Atti del XXXIII Congresso della SLI (Napoli, 28–30/10/1999), ed. Federico Albano Leoni et al. (Roma: Bulzoni, 2001), 85; Federico Albano Leoni and Pietro Maturi, “Per una verifica pragmatica dei modelli fonologici”, in La linguistica pragmatica - Atti del XXIV Congresso della SLI, ed. Giovanni Gobber (Roma: Bulzoni, 1992), 39-49. 16 Renata Savy and Francesco Cutugno, “Su alcune correlazioni tra riduzioni segmentali e tratti prosodici nel parlato spontaneo: il ruolo del fattore tempo”, in Atti del XXVII Convegno Nazionale dell’Associazione Italiana di Acustica, Genova, 26-28 Maggio 1999, 183-87. 17 Tullio Telmon, “Varietà regionali,” in Introduzione. La variazione e gli usi, ed. Alberto Sobrero (Roma: Laterza, 1993), 112-13.

Rhythmic-Prosodic Aspects in the Acquisition of Italian

45

the same speaker in both codes (Italian and dialect).18 The speaker exhibited a considerable discrepancy between sequences of duration values (which is likely to be explained by a different time structuring of the sentences in the two linguistic codes). In some cases, the interviewees also noticed the idiolectal (and diatopically marked) traits. The question, “Do you like dialect?” in some cases received the following answers: MD: It depends on the speaker. Some people have a hard pronunciation and others more melodious. MS: It depends on the person and on which part of Sicily they come from. E.g. Do you know Paolo? I understand him when he speaks dialect, but I don’t understand Ciccio. Maybe because Ciccio comes from Caltanissetta. [Paolo comes from Catania].

The pragmatic variable of variation in the intonational structure of sentences, which is linked both to the information structure and to the speaker’s emotional attitude and, consequently, to the shades of meaning attributable to communication contents, should be added to the sociolinguistic variable examined so far. Our interview included questions on meta-skills in the perception of intonational interrogation traits as well as questions suitable for revealing the real level of understanding of statements of order and information.

2.1.3. Interrogative tone and interfering diatopic variant As to the interrogative tone, the rising tone is recognised by 50% of the interviewees (i.e. MT: “you feel that it is a question”) and is judged to be the same as in their mother tongues (Spanish, Croatian, German, and French). However, in the opinion of the subjects (30%), when the interrogative tone is not clear, some help in understanding the meaning of a question where words are not understood (MT: “when a word is missing”) derives more from the context rather than from the tone. In other cases (20%), instead, the problem is not connected with intonation, but with the type of question, namely if you are being asked an open question or a closed question, as declared by a Spanish woman 18 The studies by Romano and Interlandi, however, included ways of speaking from the region of Salento (Apulia). Antonio Romano and Grazia Maria Interlandi, “Variabilità geo-socio-prosodica: dati linguistici e statistici”, in Géolinguistique, 3 (hors série: Projet AMPER - Atlas Multimédia Prosodique de l’Espace Roman, 2005), 259-80.

Chapter Two

46

during the interview: AG: Yes, I grasp that it is a question. Sometimes, but not often, I don’t understand the question. This is not due to intonation, but to the words that I am told. Based on how the question is formulated, I know that the person is asking something from me, but I don’t catch its meaning in full. That is, am I to answer yes or with a sentence?

Without overlooking our starting point (the Sicilian local variant), we wondered whether the diatopic variant could interfere also in the intonation of a question. As discovered in some of our interviews, the diatopic variant seems to interfere in the interrogative tone. Some questions included in the test (“Do you always understand when you are being asked a question? Even though you do not understand the words, do you understand that it is a question?”) confirmed that understanding a statement was strictly connected to Sicilian intonation for 40% of the interviewees. If in some cases (25%) Sicilian intonation made it rather difficult to understand the meaning-bearing rhythmic-prosodic traits, in other cases subjects (35%) considered the Sicilian interrogative tone as the only component that makes them understand the statements, as testified by some interviewees who had spent some time in other parts of Italy (Milan, Florence, etc.): PL: No! I don’t always understand when I am asked a question, I don’t understand because my ear is used to the Sicilian accent. I don’t understand Milanese and Roman ways of speaking, I need the Sicilian intonation. Before, I couldn’t understand it, but given that I learnt my Italian in Sicily, my ear is tuned to the Sicilian accent. I watch a lot of local Sicilian TV and consequently cannot understand anything else. In general, however, I understand that the tone is “rising”.

2.1.4. The diatopic variant interrelated with diastratal and diaphasical variants Together with diatopic variability, we observed the presence of a strong diaphasical or diastratal variability within the same region. This further complicates understanding statements in real situations. Indeed, 35% of the subjects declared that they find it hard to understand statements characterised by diatopically- and diastratally-marked rhythmic-prosodic traits, as appeared clearly in one answer: MT: The first months were very difficult for me: with my Italian learnt at school and the very small amount I took at the university, it was difficult ...

Rhythmic-Prosodic Aspects in the Acquisition of Italian

47

because the pronunciation typical of Sicilian ... for example, when they asked something from me at the bakery, it sounds totally different from mine; but now I’d say I understand.

In environments with a strong diastratal connotation, the typical features of how Italian is pronounced in the extreme south of Italy are still more conspicuous. We already know about consonantal reinforcement at word start or after the article and retroflection or cacuminalisation of some consonants ([l], [r], [t], [d], and [n]) or consonantal nexus.19 As detected in the analysis of data, the interference of the diatopic and diastratal variables affects acoustic phenomena (duration, frequency, intensity, spectrum configuration) and the corresponding hearing phenomena. Retroflection or cacuminalisation of some consonants, for example, is not perceived as a phenomenon of sound intensity but rather of sound length. Confirmation came in the answer of one of the interviewees who said: PL: Sicilians made words longer. For example, the sound [œ]. They told me at the market, “a[œ]petta ti do una bu[œ]ta” (Wait, I’ll give you a bag).

Mutual relationships are remarkably complicated by the many interferences. As regards the factors more directly involved in suprasegmental features, there is a closely intertwined set of mutual influences. Intensity affects fundamental frequency (and indirectly also subjective height) because a sudden increase in the speed of the transglottal airflow determines a higher frequency of vibration of the vocal cords. Duration influences intensity since it regulates the keeping of the sound energy produced in time. Finally, volume is influenced by almost all remaining parameters: by duration indirectly in as much as duration affects intensity and by fundamental frequency directly (because as vibration grows per time unit, also the feeling of sonority increases).20 The above is confirmed by the same answer of some learners-subjects who found that “at the market in Catania, they speak fast and speak in a high tone” (MT and MD). Since the same effect can be obtained through different articulatory strategies, the foreign listeners would understandably be uncertain about the strategy effectively adopted by the speakers.

19

Corrado Grassi et al., Introduzione alla dialettologia italiana (Roma/Bari: Laterza, 2003), 70-3. 20 Bertinetto, Strutture prosodiche, 33.

48

Chapter Two

2.2. The external diatopic component in the STUD Group As shown,21 it is at an intonational level that the differences among speakers of different ethno-cultural and socio-linguistic origins do not disappear. Total prosodic skills are the final stage of competence to be attained when acquiring a foreign language. This is why one of the first questions in the questionnaire was intended to highlight the problem of the tonic accent in Italian words. The answers of the study groups to the questions, “Which Italian words are difficult for you to pronounce? Why?” were homogeneous, independent of the different origin of the students. For 80% of the subjects (from Germany, Croatia, France, and Spain), the problem with Italian words most times is not a phonetic issue, but “a problem of accent”. BD [slavic subject]: We probably have no idea where to put some accents […] sometimes, I’m in trouble when I am going to say a word and don’t know if the accent comes before or after, for me both [variants] sound well.

Now, we know that the accent contributes to identifying a segmental unity in the spoken chain acting at a syntagmatic level. As far as Italian is concerned, we know that the accent performs a culminative (configurative or distinctive) function that consists in highlighting a syllable relative to the adjacent ones so as to confer “accentual unity”, a precise physiognomy or configuration.22 In addition to correct pronunciation of words containing the diphthong [au] (felt (NVM) as “unnatural”), words with the [pn] pair of consonants (e.g. pneumatic), and words with the sound [r] (single or double) (PL: “Italian [r] is not natural, I am in trouble when I’m to utter an [r]”), our French subjects found that the problem of the accent is the most important, as clearly testified by one of them: PL: In French, the accent always falls on the last syllable while in Italian you should know where it is … I don’t know if there is a precise rule about where the accent falls. 21 See John Gumperz, Discourse strategies (Cambridge: Cambridge University Press, 1982) and Voghera, Sintassi e Intonazione. 22 In Polish, for example, the accent performs a demarcation function with fixed accent and consists in limiting the width of the accentual unit itself: each lexeme is accented on the last-but-one syllable so that the mere presence of a prominence signal is enough to identify the boundary of the word. See Bertinetto, Strutture prosodiche, 42-3.

Rhythmic-Prosodic Aspects in the Acquisition of Italian

49

As often happens with fixed-accent languages, learners do not develop a perceptive strategy specific to the sound parameters. Only when they become able to master a different prosodic system do they feel the urge to make distinctions that are unusual for them. The same phenomenon occurs with the Spanish of STUD Group who exhibit problems in uttering “too long words” (in addition to difficult pronunciation of phonemes like [‫]ݹ‬ [ts] [s].  AG: I find it hard to pronounce the third person singular... For example, riuscirebbe (might manage) and pronuncerebbe (might pronounce) are words too long for me. You should make pauses within the same word, one, two, I don’t know … I wonder whether you should make pauses within the same word … long words.

As noted, pronunciation is hard due to the need to hold voice emission over the points of a very long succession of atone syllables. Interestingly, in some cases learning a second language may affect the utterance of statements in the mother tongue: PL: In Italy, you speak in a high voice; in France, the voice is kept lower. Also my mother told me so, that I had got another attitude. Since when I speak Italian, I am more “extravagante” and “speak louder also in French”, it looks like I am drunk.

In this sense, at an acoustic level, the placing of the Italian accent seems mainly intended to increase duration of the stressed vowel, which confirms how “‘tonal’ modulations of accented syllables are strictly dependent on intonation” (dialect accent).23

23

Bertinetto, Strutture prosodiche, 79.

Chapter Two

50

2.3. The external diatopic component in the IMM Group Literate in L1 and the Latin alphabet 2 Bengalis (1 with a High school diploma and 1 with a degree) 1 Mauritian with High School diploma 1 Palestinian with a degree

Literate in L1, not in the Latin alphabet 2 Mauritians 1 Moroccan

Not literate in L1 (Latin alphabet) ͳ—”‹ƒ„‡

1 Tunisian

2 Rumanians  ͳ—Ž‰ƒ”‹ƒ

3 Chinese

1 Slovenian

Not literate in L1 (nonLatin alphabet) 2 Bengalis 4 Chinese

1 Kossovarian

1 Nigerian 1 Senegalese Table 2-3: IMM Group. Our teaching experience enabled us to observe which role the external diatopic component also plays in the other control group (IMM group), and in Chinese-speaking students in particular. On one hand, explicit awareness of the tonal features that enable distinguishing and classifying an interrogative tone is acknowledged, as shown by Chinese students (belonging to the IMM group) interviewed on the interrogative tone metaskills: LD: They (Italians) pronounce the end of the word with an accent: Quanti anni hai? (How old are you?) There’s an accent on the word hai.

Whereas, on the other hand, some real cases showed differences in receiving/answering closed or open questions. For example, the IMM named Luca, an 18-year-old Chinese boy working at the Catania market. As to the rhythmic-prosodic aspect, when the student was asked closed questions (with a yes-or-no answer), he answered yes. But, when requested to provide an open answer, the student replied Non ho capito (I haven’t understood), pronounced, inter alia, as one word with no understanding of the syntagmatic boundaries

Rhythmic-Prosodic Aspects in the Acquisition of Italian

51

Teacher: Do you remember what poliziotto (policeman) means? Luca: Yes. Teacher: Can you explain the meaning of this word? Luca: [Silence] Teacher: So? Luca: I haven’t understood.

The student’s initial affirmative answer confirms what Balboni said,24 namely that a closed question with a yes-or-no answer asked by the teacher, receives a positive answer from the student (meant to confirm the teacher’s status) and that a later assessment through a more insistent question results in the unconditional “surrender” of the student. Of course, it was not laziness by the Asian boy, but respect for a European person who, asking a closed question, wanted to receive the answer, “Yes”.

3. The diamesical variation The action of linguistic literacy achievement performed by television, mainly in terms of morphosyntactic and lexical traits, is generally recognised,25 we therefore attempted to assess the influence, if any, that such a medium can exert on suprasegmental skills of L2 Italian. Among other things, this choice was also important for one of the problems connected with the analysis of intonational prosodic traits. As some have indicated, the construction of a series of synthetic stimuli imitating the acoustic properties of the spoken language poses problems, mainly when you are to build stimuli that can be subject to different perceptive classification. For this reason, in our analysis, we did not use a vocal synthesiser or a professional speaker reproducing ad-hoc built sentences, but decided to use spoken parts from television broadcasts as exempla ficta of oral texts. This phase of our investigation led us to: 1) Discover, incidentally, which are the Italian TV shows that foreigners like most (fiction, talk-shows, news and surveys on current events); 2) Identify perception of relations between tonal trend and intonational sequences specific to the different programmes (for example, the yelled hyperbolic Italian in entertainment broadcasts); 24

Paolo Balboni, Parole comuni, culture diverse. Guida alla comunicazione interculturale (Venezia: Marsilio,1999), 75. 25 See Tullio De Mauro, Storia linguistica dell’Italia unita (Roma-Bari: Laterza, 1963) and Diadori, L’italiano televisivo.

52

Chapter Two

3) Reproduce television texts in the classroom in order to exploit their intrinsic, even though involuntary, educational potential in order to support the diaphasical skills of the foreign learner. As far as point (1) is concerned, data collected showed that the most watched TV shows are news and entertainment programmes and, among them all, the favourites are “programmes with games”, i.e. Chi vuole essere milionario (The billionaire) and reality shows such as Grande Fratello (Big Brother).26 The infotainment programme Striscia la notizia is also highly popular.

3.1. The television speech collection The television texts listened to during our interviews were representative of the major macro-kinds (news and entertainment) and of a hybrid macro-kind that was selected as an example of simulated spoken language: the American soap-opera “Beautiful”, on the one hand, as an example of a serial dubbed into standard Italian; and the Italian “Un posto al sole” on the other, in marked diatopy, as example of regional Italians from Campania. The fragments of television speech, lasting two minutes each, were presented to the 15 students of the STUD group. Listening took place through loudspeakers, in a sufficiently soundproofed environment, and at a convenient volume. No special information was supplied to the students interviewed about the construction of the material. They were simply instructed to describe the rhythmic-intonational elements they perceived for each stimulus (with the assistance of dedicated questions). The selection was “compulsory”, which means that even in doubtful cases the students were required to state their preferences. In general, as regards identification of the different programme types by hearing, the answers of the interviewees were consistent in identifying the type of programme on the basis of the type of language spoken. As regards the specific types of language spoken, data were analysed as follows.

3.2. News As regards news, we selected text fragments from the main Italian newscasts. TV news spoken language can be considered as an intermediate 26

The selection of these two programmes might be explained on the basis of their international television format.

Rhythmic-Prosodic Aspects in the Acquisition of Italian

53

utterance between read speech and spontaneous speech, since such texts are written specifically for oral production. The results of some experimental surveys carried out on collections of spoken Italian that originated from radio and television show unmistakably that the rhythmicprosodic features of the spoken language used in radio and TV newscasts have deeply changed over the last fifty years. As noted,27 TV news today is no longer delivered by professional speakers, but by journalists who characterise the newscast style through pronunciation, cadence, and speaking rate (e.g. Enrico Mentana whose nickname is mitraglia, machine-gun fire, for his fast-paced speaking). The personalisation of the prosodic element has, then, allowed journalists to work out stylistic and expressive modes and solutions that clearly manifest information, suasive and perlocutionary aims and that, together with proxemic and mimic-facial attitudes, determine the “communication idiolect” of the single journalists. While selecting the news text collection, we decided to examine two different types of significant spoken language specific to this kind of programme, namely, speakers delivering news and journalists during a reportage. Furthermore, male and female journalists’ voices were included in order to give the same attention to male and female spoken language. The newscast fragments (in a monological, or mainly monological form) even though they proved not perfectly functional to the use of the phonic-hearing system since they had no speech alternation (one of the basic checks in speech production/programming), in any case enabled us to catch the tension between the dialogical primary speech condition (aimed at by all spoken speeches) and text structures suitable to support less natural communication situations.28

3.2.1. News:Collected data The questions about rhythm, rate, intonation and the relation of the rhythmic-melodic component to the scansion of news contents within statements, produced different results. As regards the speaking rate, in the opinion of the interviewees, journalists (of both sexes) speak “at an average speed” (some said, “More slowly”). 27

Edoardo Buroni, “La voce del telegiornale. Aspetti prosodici del parlato t,elegiornalistico italiano in chiave diacronica”, in L’italiano televisivo 19762006, Atti del Convegno Milano, 15-16 giugno 2009, eds Elisabetta Mauroni and Mauro Piotti (Firenze: Accademia della Crusca, 2010), 387-88. 28 Voghera, “Teorie linguistiche”, 77.

54

Chapter Two

As regards the relation of real spoken language and broadcast speech, results were unanimous and all noticed that the journalists “do not speak as you speak the real spoken language”, but then the description of broadcast speech deploys in a variegated number of types when the attempt is made to describe exactly in which ways it is “different”: a) 45% of the interviewees think that the journalist “sticks to one tone only” (AG); b) 35% that the rhythm is “straighter” (MD); c) 10% that the trend is not “too high, nor to low”, but follows a “wavy pattern” (PB); d) 10% that the journalist has a fast pace “that reveals more enthusiasm and participation in the news” (MT). As for the intonational trend, a very wide-ranging tonal variety was noted. In particular, as regards the last two answers (“wavy pattern” and “enthusiastic tone”), we noticed that the subjective impression of the speaker’s tone (high-pitched or low-pitched) may be due to prosodic facts that have nothing to do with the F0, but concern, for example, vocal emission (intensity) or the duration of the phonic segment in question. Consequently, a very energetic vowel tends to be perceived as higher than a corresponding one uttered with less energy (similarly, the voice “held” longer to pronounce a vowel of great duration, will produce a signal that, conditions being equal, will be perceived as higher than that produced by a short vowel.29 The analysis of perceptive answers relating to this sub-collection, however, made us notice how the construction of the meaning not only pertains to the tone level of the statements, but is also linked to the timing of the statement: i.e. speed is a function of text metrics. Actually, you use a different rate in the different text sections depending on whether you want to close up on certain information or whether you want to increase or decrease the listener’s attention towards that part of the text.30 As regards the language spoken in television newscasts, the interviewed learners agreed in noticing pauses that are functional to the piece of information contained in the statement, or a longer length of the

29

De Dominicis, Intonazione e contesto, VII. Francesco Cutugno and Renata Savy, “Correlation between segmental reduction and prosodic features in spontaneous speech: the role of tempo”, in Proceedings of XIVth International Conference of Phonetic Sciences (ICPhS) (S. Francisco, 1999), 471-74. 30

Rhythmic-Prosodic Aspects in the Acquisition of Italian

55

phonic segment (voice is “held” longer to pronounce a vowel with a function in the text). To the question, “Do you understand which are the most important words and key-subjects from the intonation of the sentence?”: a) 45% of the subjects noticed that the journalist makes a pause just before uttering some key-words; b) 35% realised that the journalist lengthens the most important words or makes a small pause “and key-words are emphasised by the tone” (PL); c) 20% noticed that the journalist, mainly in the titles, “prolongs vowels that should not be prolonged” (MD). As regard pauses, we make reference on the different strategic use of pauses by TV newscast speakers in recent times. The element that should be underlined is the syntactical use that is made of intonation. Actually, in most cases, immediately before the pause, there is a strongly marked intonational peak whose aim is evidently to make the listener understand that there is a break, a limit in that point.31

3.3. Fictional television shows As far as fictional television shows are concerned, we listened to a fragment of the soap opera “Beautiful” in the Italian version and a fragment of “Un posto al sole”, an Italian production, where the diatopic mark is very clear. With regard to “Beautiful”, it was noticed that the speaking rate is slower than in spoken language: a) 80% of the interviewees declared that the actors “speak slowly and clearly” (MT); b) 20% declared that the actors speak more slowly than in current spoken language, “almost whispering”, but there are also “intonational exaggerations”. It is interesting to observe how the relationship between syntax and intonation is perceived as being “grammatically correct”, namely “you feel the concordance of times” (PB).

31

Giannini and Pettorino, “Il parlato dei mass media”, 76-8.

56

Chapter Two

Listening to “Un posto al sole” immediately after “Beautiful” prompted the learners to compare the speed of the fragments proposed. Data collected show that: a) 75% of the subjects declared that the Italian production features a more natural, real, familiar speed. “This is an Italian production, isn’t it? You realise that is more natural” (MD); b) 25% found mitigated evidence of simulated speech given that “they try to speak at a more natural rate, not as when reading” (MT); c) 10% noticed a mode of speaking more similar to real spoken language, and at the same time faster than in “Beautiful”. We have already observed that “Un posto al sole” is an interesting specimen among the exempla ficta, as its plot ranges over a number a variants such as broadcast speech, simulated speech and diatopically marked speech in particular. In one of the fragments listened to, interviewees noticed codeswitching between standard Italian and regional Italian, also in connection information therein. Indeed, they pointed out the following: MD: You listen to a “grammatically correct” Italian when they talk about something important (such as the house), in other cases things change and there are “parts” where they speak dialect. You understand this not from the words, but from their tone because when they are talking about the house, the tone is “slow and serious” while when they are talking of “their own affairs” (family subjects) the tone is fast and less clear and words are come out muddled.

3.4. Entertainment As for entertainment, two texts were made available. The first one is a fragment taken from “Isola dei famosi” (the Italian edition of the programme “Celebrity Survivor”), while the second is taken from the show “Che tempo che fa”, a talk show with interviews with Italian and international guests, in the style of the overseas “Late Show”. As regards “Isola dei famosi”, the fragment focussed on an on-screen quarrel between two persons (a showgirl, Simona Ventura, and a writer, Aldo Busi) and the dialogic situation they create is characterised by speech overlapping. This affects understanding of single words by the foreign learners who, however, grasped the general meaning thanks to the tone of

Rhythmic-Prosodic Aspects in the Acquisition of Italian

57

the two quarrellers.32 Data collected show that 75% of the learners “did not understand all the words, but understood the tone of the showgirl”. In any case, they perceived the simulated hyperbolic spoken language. One of the subjects said: PL: The showgirl is theatrical, wants to stage a show, stresses all tonic accents, emphasises the most important words: “I’m not in for polemics”, but from the tone of her voice tone she means “I do want to polemicize”, betrays the opposite. She didn’t want to thank him (as she says), but to polemicize.

With regard to the programme “Che tempo che fa”, we selected a fragment from an interview by the conductor Fabio Fazio with actor Riccardo Scamarcio. In this case, there is no word overlapping. On one side subjects also perceived the different orthoepic pronunciation of the actor interviewed: MT: You notice “closed o” [i.e. incontro (meeting)], you hear the difference with other people. The same applies to Fabio Fazio (the interviewer) but to a lesser even if still noticeable extent.

On the other hand, they also noticed closeness to a more genuine, lesssimulated spoken language: MT and MD: The language is more “spoken-like”, more natural as if the two of them met in the street and were having a chat; it doesn’t sound like a television broadcast.

In some cases, however, perception of the speaking rate or of prosodic boundaries changed from one subject to another depending on the perceptive judgements of the foreigner/listener. For example, perceptive identification of prosodic breaks and the strong perceptive relevance of this trait determined the observation of different talking speeds in the interviewer and the interviewee: 32

In shows like “Isola dei famosi” where participants quarrel on screen and overlap speaking turns, it is difficult for the foreign learner to grab the single words even though the general subject of the speech is understood. The question “Do you understand what they are talking about?” received answers like: “I do for the general subject, but not the words” (Mt and MD). MT: “I realise that she feels offended; her voice is shrill, almost quivering”. “He does something defensive: zac, sentences become shorter (to emphasize certain aspects) and almost jerky”.

58

Chapter Two MD: The host talks faster while the actor talks slower.

According to another interviewee: MT: The actor speaks normally while the host makes “sort of undulations”: two words slowly (at the beginning of the question) and five words rapidly.

4. Conclusions To conclude, we underline that collecting a huge amount of data relating to geographical, social and situational variability of prosody is now the objective of numerous projects over diversified linguistic areas. The already sufficiently complex cultural situation is becoming increasingly complicated due to the appearance on the horizon of a new standard of Italian that seemingly comes as a model originating in the language spoken in real life. In past years, we were certain of targeting a standard language (conveyed by literature, taught in the schools, etc.). Today, new variants referred to as average use Italian or neostandard Italian come to light that describe and define pan-Italian traits at the intonational-prosodic, phonetic, morphological, lexical, syntactic and textual levels, widely diffused in all social strata and pervasively conveyed by mass media. Specifically, as far as our corpus is concerned, one can observe how Sicilian phonetic traits seem to significantly affect the rhythmic and intonational aspects where, for example, a vowel or a consonant pronounced with greater intensity may be perceived as more intense (i.e. bearing an accent). The use of television texts for teaching purposes allows delocalisation of standard Italian and overcoming of the regional variant as to rhythmicprosodic and phonetic traits. In particular, the diamesical and diatopic variation axes can be opportunely used to reach the standard. The diamesical variant is a means to implement the standard; the diatopic variant seems to lead far from it. How does the diamesical variant influence the learning of the rhythmic-prosodic traits of standard Italian? Making recourse to a known formula, we may say that “speed = space/time” where speed is the learning speed, space is the geographical space (diatopic variant) where the foreign learner lives and, at the same time, the virtual space filled with the use of the means of communications in the teaching system; time is the length of stay of the non-native students (generally more limited for exchange students than for immigrants). Obviously, suggesting a formula should not be considered a solution, but a spur for further consideration in this field.

CHAPTER THREE PRAGMATIC MARKEDNESS IN L2 ITALIAN: SOME DATA ON THE MORPH-SYNTAX/ INTONATION INTERFACE IN MOROCCAN ARABIC SPEAKERS’ INTERLANGUAGE LAURA MORI LUSPIO UNIVERSITY OF ROME, ITALY

1. Introduction In Mori (2007) a survey was carried out on L2 Italian of Moroccan speakers in order to highlight main segmental features affecting this variety;1 phonetic-acoustic properties of vowels and consonantal variants were examined through instrumental speech analyses. The perceptual salience of these markers was subsequently tested by native Italian speakers and the foreign accent factor evaluated. A preliminary investigation into Moroccan Arabic-accented Italian also revealed the presence of non-native variants at levels beyond the phonetic-phonological dimension. In the current paper this interlanguage2 will be further described from the aspect of the morph-syntax interface. It is well-known that the morphsyntax of any spoken language differs from the grammatical rules binding written varieties, presenting a greater inner variability and complying with pragmatic-orientated strategies (the so-called “grammar of speech”). The above statement is especially valid in the case of learning varieties of an L2, especially during early acquisition stages when speakers’ 1

Laura Mori, Fonetica dell’italiano L2. Un’indagine sperimentale sulla variazione nell’interlingua dei marocchini (Roma: Carocci, 2007). 2 Larry Selinker, “Interlanguage”, International Review of Applied Linguistics in Language Teaching, X/3 (1972): 209-31.

60

Chapter Three

utterances comply mainly with pragmatic needs and grammatical competence is reduced. In fact, according to Gìvon,3 the “pragmatic mode” characterizes initial phases of acquisition, followed by a progressive development of lexical-semantic competence; morphological and syntactical competences being acquired thoroughly only in advanced stages.4 Therefore, in interlanguages, at any level, it is possible to find evidence of a path of acquisition, going from pragmatics to syntax, and note cues that substantiate this directionality. As for L2 Italian morph-syntax, Chini5 observed three main features referring to the above-mentioned acquisitional phases: the pragmaticlexical stage is marked by a massive presence of nouns, noun phrases and anaphoric references and consequent acquisition of the morphological competence when demonstratives and stressed pronouns are acquired. Finally, syntactic competence in Italian is reached when agreement, anaphoric zero and clitic clusters start to occur in L2 speech. Although many findings confirm this developmental path in L2 acquisition (from pragmatics to syntax) of Italian structures,6 it is worth noting evidence of pragmatics-based features proving the relevance of communicative needs at any stage of L2 speech. As regards L2 Italian of Moroccan migrants, a peculiar morph-syntactic pattern was pinpointed: the use of non-standard averci (to have got) with a possessive-presentative function, preceded by the accusative clitic lo (i.e. ce l’ho tempo [‫ݹ‬e‫ޖ‬l‫ޖܧ‬t‫ܭ‬mpo]. In the following pages we intend to show that this non-native construction has to be considered not only as a structurally marked feature, but also (and maybe, mainly) as a pragmatically marked construction in L2 Italian productions. In order to substantiate our pragmatic-orientated interpretation of L2 Italian acquisitional data, we are going to refer to the contribution of morph-syntax/intonation interface in detecting Moroccan-accented Italian peculiarities.

3

Talmy Givon, Discourse and syntax (New York: Academic Press, 1979). See studies developed within the framework of the European Science Foundation project. As for the acquisition of L2 Italian, see papers by Cecilia Andorno, Giuliano Bernini, Marina Chini and Anna Giacalone Ramat, of the so called “Pavia Research Group”. 5 Marina Chini, Che cos’è la linguistica acquisizionale (Roma: Carocci, 2005). 6 See Anna Giacalone Ramat, ed., Verso l’italiano. Percorsi e strategie di acquisizione (Roma: Carocci, 2003). 4

Pragmatic Markedness in L2 Italian

61

2. The sample of investigation The current research study focuses on sintagmatic contexts that contain a superficial right-dislocation: the cataphoric pronoun cluster ce lo,7 followed by the inflected form of avere (to have), that pre-announces the sentence comment. Within the Italian sociolinguistic repertoire the verb averci is a variant of avere, frequently attested in informal registers of spoken Italian. This form occurs in both popular and colloquial Italian,8 which are the most frequent input varieties for migrants acquiring Italian in a natural context. Needless to say, a speaker's selection of the ce lo cluster could be enhanced by high frequency of averci (to have got) to express possession, and of a near-omophonic form such as c’è with a presentative meaning, both occurring in informal spoken Italian.9 The occurrences of ce lo + avere construction was considered in L2 Italian speech productions by twenty-eight Moroccan speakers. The reference corpus, ArabIt,10 comprises speech material elicited through different tasks (word list, visual story-telling, free speech and guided conversations). The sample includes 13 male and 15 female speakers, with highly differentiated socio-cultural profiles, who have been living in central Italy for a time span ranging from 2 to 21 years. Linguistic data discussed below are extracted from semi-spontaneous conversations in which everybody participated, answering questions11 on different topics: intercultural differences, social inclusion in Italy, linguistic behaviour, communicative practices and their migration project. 7

In this phonetic context a phenomenon of vocalic sandhi occurs, that is to say, the elision of an unstressed final vowel whenever a vowel is the first segment of the following word within the same syntagm. Together with this external assimilation a lengthening of the preceding consonant might be observed, in this case the lateral alveo-dental /l/. 8 Gaetano Berruto, Sociolinguistica dell’italiano contemporaneo (Roma: Carocci, 1987). 9 See Jürgen Meisel, Harold Clashen and Manfred Pienemann, “On Determining Developmental Stages in Natural Second Language Acquisition”, Studies in Second Language Acquisition 3/2 (1981): 109-35; Monica Berretta, “Marcatezza in morfologia e apprendimento di lingue seconde”, Quaderni del Dipartimento di Linguistica e Letterature Comparate 8 (1992): 129-56; Monica Berretta. “Morphological markedness in L2 acquisition”, in Iconicity in Language, ed. Raffaele Simone (Amsterdam: John Benjamins, 1995), 197-233. 10 Laura Mori, Fonetica dell’italiano L2. Un’indagine sperimentale sulla variazione nell’interlingua dei marocchini. 11 Conversations within the ArabIt corpus were led by myself, as interviewer.

Chapter Three

62

During an early phase of analysis we inspected frequency and distribution of the above-mentioned construction in Moroccan-accented Italian; its presence was found to be strictly correlated with the speakers’ acquisitional stage. In fact, the use of ce lo pronominal cluster to refer to a cataphoric element is not attested in initial stages of acquisition when the possessive-presentative c’è12 is overextended. In non-native Italian, pronominal clusters in general are attested in initial post-basic varieties13 once speakers have crossed the basic stage of L2 acquisition. The correlation with the acquisitional stage implied a sample reduction: out of 28, only nine speakers presented features of an initial/intermediate post-basic variety of L2 Italian.14 Among these, occurrences of the target right-dislocated construction were attested in Moroccan-accented Italian by four speakers: two men (1MRM; 9MVT) and two women (2FVT; 8FVT) with quite different backgrounds. In tables below, information on each speaker’s socio-cultural profile and linguistic biography are reported. Sex Town of origin Italian residence Age of arrival Length of stay in Italy Education Job Language(s) at home Language (s) at work

Male Kouribgha Rome 30 5 years High school (+ 2 yrs. university classes) Pastry chef Egyptian Arabic Italian

Table 3-1: Speaker 1MRM.

12

It is interesting to remark that in the contact variety of Italian developed in Ethiopia (Simplified Italian of Ethiopia) the relationship of possession can be expressed by “ce”, as alternative to the use of the verb to have (aßere); i.e. “Ѡyo non ce makkina” (non ho la macchina). See Marcos Habtemariam, “Italian”, in Language in Ethiopia, eds. Marvin L. Bender, Donald Bowen, Robert L. Cooper and Charles A. Ferguson (Oxford: Oxford University Press, 1976), 170-182. 13 Wolfgang Klein and Clive Perdue, “The basic variety (or: Couldn’t Natural Languages be simpler?)”, Second Language Research 13 (1992): 301-47. 14 Data from different linguistic levels were considered in order to classify speakers’ linguistic competence according to the learner’s variety categorization. See Giacalone Ramat, Verso l’italiano.

Pragmatic Markedness in L2 Italian

Sex Town of origin Italian residence Age of arrival Length of stay in Italy Education Job Language(s) at home Language (s) at work

Female Kouribgha Viterbo 28 4 years High school Housemaid Italian -

Table 3-2: Speaker 2FVT. Sex Town of origin Italian residence Age of arrival Length of stay in Italy Education Job Language(s) at home Language (s) at work

Female Casablanca Civita Castellana (VT) 21 18 years Primary school Bar owner Moroccan Arabic Italian

Table 3-3: Speaker 8FVT. Sex Town of origin Italian residence Age of arrival Length of stay in Italy Education Job Language(s) at home Language (s) at work Table 3-4: Speaker 9MVT.

Male Rabat Civita Castellana (VT) 25 2 years Junior high school Factory worker Moroccan Arabic and Italian Moroccan Arabic and Italian

63

64

Chapter Three

3. A peculiar morph-syntactic construction: “ce lo + avere + Obj.” Andorno15 had attested occurrences of this dislocated structure in L2 Italian by Moroccan adolescents, interpreting this feature as an evidence of univerbation due to a morphological difficulty in word segmentation. In her opinion, learners of Italian consider the clitic cluster ce lo as inseparable from the verb, especially in the case of verbs such as avere, where this agglutination determines a phonic increase of inflected verb forms. Listed below are the sentences extracted from ArabIt corpus16 (speakers 1, 2, 8, 9) where the pronominal cluster ce lo is attested in preverbal cataphoric position, preceding the sentence-comment encoded in the following direct object. This morph-syntactic configuration is relatively unusual if compared with native-Italian, where pronouns or pronominal clusters are used to refer anaphorically to a referent previously activated within the text. 1a)

S: Sai quan[t]o arrivato a Italia prima...me prima stato un mese a Roma, dopo un mese a nord. A nord c’è lavoro me purtroppo sen[s]a documenti non te fanno lavora’ me, lavori solo con [l]i stranieri come te, ca[b]ito? che ce l’hanno dei pro[‫]ݤ‬et17.

1b)

I: E gli amici che abitano con te lavorano con italiani? S: uno ce l’ha una pizzeria, lavora con su fratello. 18

1c-d)

S: E ritornato a Roma, prima vendo li [bi]stiti, vendo li [bi]stiti, dopo ce l’ho l’affitto.... ce l’ho tanti cosa.19

15

Cecilia Andorno, “Prima parla poi pensa: successo di una strategia di acquisizione basata sulla copia in Italiano L2”, Studi italiani di Linguistica teorica e applicata XXV/2 (1996): 291-311. 16 The following speech samples are extracted from semi-guided conversations between the interviewer (I) and the speaker under examination (S). Our English translation of these exchanges, maintaining source language solecisms, is given in footnotes. 17 “S: When I arrived in Italy I first spend one month in Rome, afterwards I moved to the North. In Northern Italy there is job but if you don’t have regular documents you are not allowed to work, you can work only with other foreigner as you, those who have got projects”. 18 “I: What about the job of the friends living with you? S: One among them has got a pizzeria, he works with his brother”.

Pragmatic Markedness in L2 Italian

65

1e)

I: Sei venuto in Italia anche per studio? S: no, no in Italia venuto solo per lavoro. Ai ce l’ho neputi mia adesso finito ma purtroppo non fa niente.20

2a)

I: Abitualmente esci con Italiani oppure esci con persone del tuo Paese? S: io esco di più con le persone Italiana che d[i] d[i]l mio Paese.... I: Sì? S: Sì. I: Però con altri Marocchini hai ancora contatti o... S: No, ce l’ho però meno perché me avuto tanti problemi allora non ce l’ho tanti... non ce l’ho tanti... come si dice? Prima ce l’ho tanti... del mio Paese... però trovato tanti problemi allora... adesso ce l’ho l’amici italiani di più del mio Paese21

2b)

I: Non hai mai fatto comunque un corso d’italiano? S: No, no. Un corso no, vorrei farlo. Sono scritta però... questa cosa difficile per noi. S[i] uno viene qua viene per lavoro, studio s[ig]onda cosa... non c’è una fami[lj]a per dir[i] se ‘c’ho la mia famil[ja] allora ce l’ho il tempo per andare a studiar[i]22

8 a-b)

I: Però ti dispiace? P7: Sì mi di[ȉ]piace sì, ma [b]urtroppo dove sta il pane dicano sta... quello [b]ure è il tuo paese, [b]urtroppo però ti manca la

19

“I: And once back to Rome, at the beginning I sold clothes, I sold clothes, then I have got rent, I’ve got a lot of things (to pay).” 20 “I: Did you come to Italy to study too? S: No, not at all. I just came to work. I’ve got a niece of mine who has finished (her studies) but now she’s finding anything to do.” 21 “I: Are you more used to meet with Italians or with your connationals? S: I’m more used to go out with Italians rather than with people from my Country. I: Really? S: Sure. I: But are you still in contact with other Moroccans or not? S: No, I still have contacts but fewer because I had many troubles, that’s why I don’t have many, I don’t have many... how do you say? At the beginning I had a lot coming from my Country... but I got into troubles with them so now I’ve got more Italian friends than (people) from my Country.” 22 “I: Have you ever attended any Italian language course? S: No, no. Not a language course, I’d like to attend it. I enrolled but... this is difficult for us. If you come here to work, studying is secondary... You don’t have your family here, if I had my family here I would have time to go to school.”

66

Chapter Three

natura, la fami[lj]a perchè li fi[j]i sono nati qua non può, so’ cresciuti qua uno ce l’ha... la masch[ԥ] ce l’ha undici anni e la femmina ce l’ha sei.23 8 c)

I: Senti, e quindi adesso hai più contatti con Italiani o con...? S: No, io so’ aperta per tutti, non c’ho problema... I: Ma amici? S: Amici...perché quelli che c’ho so’ qua... per qua... quello lavoro mio ce l’ho l’amici italian[e] di più dal mio paesani.24

9 a-b)

I: non conoscevi nessuno che abitava già qua? S: sì, ce l’ho...siamo qui... cugini, ca[b]ito? tutti cugini, va bene così. C[i] l’ho quattro cugini... ce l’ho una zia fuori da Civita.25

Sentence configuration of the above-mentioned examples presents a canonical order with the topicalized subject in initial position, complying with Italian as a topic-prominent language. Subjects coincide with the speaker himself/herself in all cases, except for sentences 1a, 8a and 8b, where the informative role of topic is hold by a third-person grammatical subject. More precisely, the subject is omitted, being a highly accessible referent: either the first person singular pronoun (the informant himself/herself) or the third person singular or plural personal pronouns (somebody recently activated within the text). The peculiarity is represented by the pre-verbal presence of the ce lo cluster where the direct object pronoun lo doesn’t refer to anything traceable to the cotext, nor to anything to be inferred from the situational context or cultural background. In fact the structure under examination presents a reduplication in discourse: the clitic lo being a syntagmatic

23

“I: Do you feel sorry for that? S: Yes I do, but the place you find your living turns out to be your Country unfortunately, but you miss your homeland nature, your family. My sons were born here, they have grown up here, my son is eleven and my daughter is six years old.” 24 “I: What about now, are getting more in contact with Italians or with…? S: No, I’m open-minded, I don’t have any problem at all. I: What about your friends? S: My friends live here, I have got Italian friends at work, more than Moroccan ones.” 25 “I: Did you already know somebody living here, before moving to Italy? S: Yes, here there are my cousins, do you see? All cousins, that’s fine. I have got four cousins … I’ve got my aunt, she’s not living in Civita (Castellana).”

Pragmatic Markedness in L2 Italian

67

anticipation of the comment encoded in the right-dislocated phonological word. Within these sentences the direct object in postverbal position represents a verb argument of avere and encodes new information concerning the sentence topic. According to this hypothesis the clitic cluster, grammaticalized as a verb agreement-marker, plays a pragmatic function because it emphasizes the informative status of the comment.26 Our examples confirm Andorno’s morphological interpretation of this marked construction as the result of a process of morphological reanalysis. As far as the informative side is concerned, Andorno27 acknowledged no specific informative value in dislocated elements.28 On this last point, we hypothesize that this morphosyntactically marked construction could rather play a pragmatic function in L2 discourse. Considering that marked sentence structures generally correlate with prosodic prominence strategies29, the instrumental analysis of morph26

In Italian—and in Romance languages in general—right dislocation has shifted to be a grammaticalized structure that emphasizes the rhematic element rather than the theme, thus losing a lot of its markedness (see Simone Raffaele, “Une interprétation diachronique de la “dislocation a droite” dans les langues romanes”, Langue Française 15 (1997): 48-61. 27 Cecilia Andorno, “Prima parla poi pensa”. 28 Ibidem: “Io non ce l’ho il quaderno, prendo una nota. Anche tante volte in classe non ce l’ho libro” (I haven’t got the exercise book, I took a note. Many times in class I haven’t got the book too); “i secondo disegno, ce l’ha la pancia grande, quello no. E questo ce l’ha le mani...piccoli, questo ce l’ha le mani grande” (in the second drawing he/she has got a big belly, not that one. And this one has got small hands, this one has got big hands). 29 See the following studies: Amedeo De Dominicis, Intonazione e contesto (Alessandria: Edizioni dell’Orso, 1992); Cinzia Avesani and Mario Vayra, “Costruzioni marcate e non marcate in italiano. Il ruolo dell’intonazione”, in Il parlante e la sua lingua. Atti delle X giornate del Gruppo di Fonetica Sperimentale, ed. Donatella Locchi (Napoli: Istituto Universitario Orientale, 2000): 1-14; Cinzia Avesani et al., “Accentazione, deaccentazione e struttura informativa”, in Misura di Parametri. Aspetti tecnologici ed implicazioni nei modelli linguistici. Atti del Primo Convegno Nazionale AISV, ed. Piero Cosi (Brescia: EDK Editore, 2005): 287-312; Mara Frascarelli, “Dislocation, Clitic Resumption and Minimality: A Comparative Analysis of Left and Right Topic Constructions in Italian,” in Romance Languages and Linguistic Theory, eds. Reineke Bok-Bennema et al. (Amsterdam / Philadelphia: John Benjamins, 2004) 99-118; Giuliano Bocci. Aspetti teorici e analisi sperimentale della focalizzazione contrastiva nella periferia sinistra della frase: sintassi, fonologia e fonetica, Tesi di laurea non pubblicata (Siena: Facoltà di Lettere e Filosofia, 2003-2004); Claudia Crocco, “Topic accent and prosodic structure”, in Information structure and its

68

Chapter Three

syntax/intonation interface could provide relevant facts. Therefore, in order to check the informative value of the morph-syntactic construction under investigation, we intend to inspect its intonational pattern.

4. The morph-syntax/intonation interface Among the re-standardization trends observed in contemporary Italian, Berruto30 mentioned a number of interesting phenomena that shape marked sentence structures through different devices: topicalization (such as left and right dislocations), c’è with a presentative function, hanging topic, and cleft-sentences.31 These marked constructions have been analysed especially as far as LD (left-dislocation) is concerned, while RD (right-dislocation) has been scarcely investigated, except for linguists’ interest in providing labels to define this latter multi-functional phenomenon: afterthought, antitopic, defocusing, detachment, extra-position, inverted word order, tail, etc.32 Whereas afterthoughts and self-repairs are right dislocation referring to a given topic, the so-called antitopic consists of a right dislocation with an emphatic-expressive meaning and a precise informative value. As regards prosodic correlates of dislocation, afterthoughts are generally separated from the finite verb by a break in time, while

interfaces, ed. Lunella Mereu (Berlin / New York: Mouton de Gruyter, 2008), 1549. 30 Gaetano Berruto, Sociolinguistica dell’italiano contemporaneo. 31 According to Rossi 2000, the most detailed study on dislocation in Italian continues to be Gossen (1954). Here left and right dislocations are analysed—as far as the intonational, pragmatic and syntactic aspects are concerned—in Italian narrative and dramatic texts dating back to the nineteenth and twentieth centuries. See Carl Th. Gossen, Studien zur syntaktischen und stilistischen Hervorhebung im modernen Italienisch (Berlin: Akademie-Verlag, 1954); Fabio Rossi, Tratti pragmatici e prosodici della dislocazione a destra nel parlato spontaneo LABLITA, 7 (Firenze: Università di Firenze, 2000). 32 See Knud L. Lambrecht, Topic, antitopic and verb agreement in non-standard French (Amsterdam: John Benjamins, 1981), 75; Knud L. Lambrecht, Information structure and sentence form: Topic, focus, and the mental representation of discourse referents, in Cambridge Studies in Linguistics 71 (Cambridge: Cambridge University Press, 1994), 202-3; Chafe Wallace, “Givenness, contrastiveness, definiteness, subjects, topics, and point of view”, in Subject and Topic, ed. Charles N. Li (London / New York: Academic Press, 1976), 25-56.

Pragmatic Markedness in L2 Italian

69

antitopics are incorporated along with the finite verb as part of a unified contour, with neither pauses nor pitch excursions.33 In sentences by speakers 1, 2, 8, 9 we remark a constant syntactic informative pattern: Speaker 1.a 1.b 1.c 1.d 2.a 2.b 8.a 8.b 9.a 9.b

Subject34/Given (gli stranieri) (lui) (io) (io) (io) (io) il maschio la femmina (io) (io)

ce lo + V (avere) ce l’hanno ce l’ha ce l’ho ce l’ho ce l’ho ce l’ho ce l’ha ce l’ha ce l’ho ce l’ho

Object/New dei projets pizzeria l’affitto tanti cosa l’amici italiani il tempo undici anni sei (anni) quattro cugini una zia

Table 3-5: Sentence informative structure. According to Berruto,35 in spoken Italian, right-dislocation of a noun/noun phrase can occur “outside the sentence”, albeit anticipated by a copy-pronoun referring to it.36 Quite interestingly he interprets these cases in terms of an extraposition of the VP (verb phrase) where the new information is encoded. In the above-mentioned sentences from our corpus, objects included within the sentence VP introduce a new item of information within the discourse, and this informative status could be enhanced prosodically through pitch excursion. In order to check the potential pragmatic value of this morph-syntactic structure, an intonational analysis was carried out aimed at describing its melodic contour; first of all, speech stretches were 33 See Lambrecht, Topic, antitopic and verb agreement in non-standard French; Gaetano Berruto, “Le dislocazioni a destra in italiano”, in Theme-Rheme in Italian, Thema-Rhema im Italienischen: symposium. Tema-Rema in Italiano, ed. Harro Stammerjoann, (Tübingen: Günter Narr Verlag, 1985), 55-69; Mario Sala Gallini, “Lo statuto del clitico nella dislocazione a destra: pronome vero o marca flessionale? ,” Archivio glottologico italiano LXXXI/1 (1996): 76-94. 34 In some sentences the subject is inferred, in this case we put the implied reference into brackets. 35 Gaetano Berruto, “Per una caratterizzazione del parlato: l’italiano parlato ha un’altra grammatica?” in Gesprochenes Italienisch in Geschichte und Gegenwart, eds. H. Holtus Günter et al. (Tübingen: Günter Narr Verlag, 1985), 120-53. 36 Such as le mangio le mele (*I eat them, the apples).

70

Chapter Three

analyzed using Praat (5.2.10 version) in relation to the main phoneticacoustic parameters, such as F0, intensity and segmental duration. The prosodic labelling we referred to is based on INTSINT (INTernational Transcription System for INTonation), an annotation system that interprets the intonational contour as a sequence of tonal targets codified according to their relative pitch. As for fundamental frequency extraction, the algorithm MOMEL (MOdélisation de la MELodie) was applied to the automatic derivation of targets points. This method of intonational analysis is based on a hybrid phonetic/phonological model: on the basis of a low-level phonetic analysis technique (MOMEL), a phonological description is provided using INTSINT.37 The series of target points determined by Momel modeling technique represents the phonetic input, which is then described phonologically referring to a set of INTSINT tonal symbols: T (top), M (mid), B (bottom), H (Higher), S (Same), L (Lower), U (Up-stepped), D (Down-stepped). In the following graphs, a representation of pitch movements is given for each of the sentences presenting the phenomena under investigation: the three rows below the F0 contour contain the following information— phonetic transcription of the string (row 1), F0 values (row 2) and INTSINT tonal labelling (row 3). Intonational analyses shows a uniform F0 profile in correspondence with the ce lo + avere construction, thus confirming that the sequence, morphologically re-analyzed, is treated as a single phonological unit within which the new information is encoded. As a matter of fact, our analyses spotlight either a rising movement of the pitch on the stressed syllable of the phonological word referred to by the direct pronoun or a pitch increase in its correspondence (prosodic labels T or H). In five of the above-cited examples, NP heads (1a: projets, 1b: pizzeria, 1c: affitto, 1d: cosa; 2b: tempo) are prosodically focalized through pitch resetting. In 2a and 8c the adjective italiani (Italian), referring to friends, is prosodically marked in focus position; in 8b the pitch rising coincides with numeral referring to the age of the speaker’s daughter: undici (eleven). Finally a contrastive focus is attested on the adjective mia (my) in 1e, in order to highlight the possessive relationship between the speaker and his niece. 37

Daniel Hirst, Albert Di Cristo and Robert Espesser, “Levels of representation and levels of analysis for intonation”, in Prosody: Theory and Experiment, ed. Merle Horne (Dordrecht: Kluwer Academic Publishers, 2000), 51-87.

Pragmatic Markedness in L2 Italian

71

In general, the attested pitch movement alignment with the sentence focus contributes to signaling the new informative status of the phonological word; in some cases the pragmatic value of this interface is enhanced by the anticipatory pitch rising on the ce lo + avere cluster (see Figures from 3-6 to 3-17).

5. Conclusions This preliminary study on a sample of Moroccan informants revealed occurrences of the morph-syntactic structure ce lo + avere + Obj. in L2 Italian speech in four out of nine speakers with an initial/intermediate post-basic competence. The current objective was to collect prosodic cues for a better understanding of the functional value of the non-native use of this morphsyntactic pattern. The most plausible interpretation seems to be its informative function as a morph-syntactic/intonational frame for pragmatic emphasis. The intonational evidence substantiates the following hypotheses: 1) The clitic cluster ce lo is re-analysed by non-native speakers as a verbal agreement-marker; 2) Speakers tend to use this marked construction to emphasize the informative function of the element in focus position. As for the first statement, this complies with Berretta’s interpretation of the Italian clitic system as a set of agreement-markers, rather than forms for anaphorical reference. She maintains, in fact, that many cases of pronominal redundancy in Italian should not be interpreted in terms of dislocation. They express a kind of “objective conjugation”: formally unstressed pronouns (or pronominal clusters) represent a single phonological word together with the inflected verb; functionally they mark agreement between verb and its object.38 Considering the morph-syntax/prosody interface, intonational contours reveal a unitary profile for this unsegmented construction where the agreement-marker cluster plays a pragmatic role within discourse (see our second statement). In fact, our informants do not generally use this construction of “superficial dislocation” to maintain cohesion within their discourse by referring to a given topic, either backwards or forwards. Speakers exploit this morph-syntactic pattern to introduce new 38

Monica Berretta, “Tracce di coniugazione oggettiva in italiano”, in L’italiano tra le lingue romanze, eds. Fabio Foresti, Elena Rizzi and Paola Benedini (Roma: Bulzoni, 1989), 125.

72

Chapter Three

information in their discourse, presenting something so far unactivated within the text and which is not easily inferred from the situational context or cultural background. As a matter of fact, the pronominal anticipation of the object through the cathaophoric lo is only superficially a textual reference; the pronominal cluster is rather inserted to emphasize the possessivepresentative relation between the verb and its object. According to this interpretation, this non-native morphosyntactic sequence would play primarily as a pragmatic device, rather than being a right dislocation for internal cohesion. Furthermore, this superficial right-dislocation could comply with a speaker’s communicative intention, as a non-native-speaker device to try to hold the interlocutor’s attention and maintain discourse cooperation. As Berruto39 points out, this kind of dislocated structure can be considered as anti-topic in terms of right-dislocation of a given element and parallel emphasis to the VP. In these cases, the giveness of the topic is not motivated by an effective cotextual or contextual reference, but is rather considered as discourse-motivated. In fact, the speech grammar is egodependent, so what is central in a speaker’s speech can be considered as an informative ‘given’ element. To this, at least in the case of our study, I would add that this morphsyntactic/intonational frame put emphasis on discourse from both sides: it highlights both the “giveness” of the topic (superficially anti-topic) for the speaker but, at the same time, its degree of ‘newness’ for his/her interlocutor. In this regard, another aspect to be investigated in the future is the potential correlation with the variable “speech typology”. As far as we know, the sequence ce lo + avere + Obj. occurs exclusively in conversational interactions, when speakers comply with the cooperative principle and, at the same time, negotiate social roles and their relationship within a speech event. In order to give further relevance to our findings, it will be interesting to analyze L2 Italian by different ethnic groups, from a morphosyntaxprosody interface position. Preliminary evidence of a similar non-native use in L2 Italian speech has already been collected in a corpus collected by the University of Pavia (Banca Dati di Pavia),40 From this corpus we extracted some occurrences (see Appendix at the end of the paper) which 39

Gaetano Berruto, “Le dislocazioni a destra in italiano”. The corpus comprises recordings of twenty-two speakers with a different competence in Italian. For details see Giuliano Bernini, “La banca dati del ‘Progetto di Pavia’ sull’italiano lingua seconda”, Studi Italiani di Linguistica Teorica e Applicata XXIII/2 (1994): 221–36. 40

Pragmatic Markedness in L2 Italian

73

show the feature to be not just a peculiar variant of Moroccans speaking Italian as a second language. In fact, the pronominal cluster agglutination construction with the verb avere was clearly attested in first post-basic stages of three informants with a different L1: a Tigrina speaker and two Chinese speakers.41 Results obtained from the intonational domain highlight the pragmatic markedness of this morph-syntactic pattern. In general, L2 Italian speakers (not only Moroccan ones) decide whether to use this marked pattern or not,42 thus confirming a kind of morphological control over its use which complies with the speaker’s pragmatic purpose. In conclusion, the application in L2 research of a protocol based on the morph-syntactic/intonation interface allowed us to corroborate the role of pragmatics in interpreting interlanguage. In our opinion, the relationship between L2 forms and their pragmatic function is pivotal and the functionalist approach proves to be the most effective in interpreting L2 outputs,43 independently from structural levels involved and speaker’s second language acquisitional stage.

41 My thanks to Cecilia Andorno for sharing speech material belonging to the corpus Banca Dati di Pavia. 42 Speakers who use the construction under investigation reveal competence in using the Italian inflected verb avere without any agglutination process with pronominal clusters. 43 Russell Tomlin, “Functionalism in Second Language Acquisition”, Studies in Second Language Acquisition XII (1990): 155-77.

74

Chapter Three

Fig. 3-6: (che) Ce l’hanno dei projets.

Fig. 3-7: Ce l’ha pizzeria.

Pragmatic Markedness in L2 Italian

Fig. 3-8: Ce l’ho l’affitto.

Fig. 3-9: Ce l’ho tanti cosa.

75

76

Fig. 3-10: (ai) Ce l’ho neputi mia.

Fig. 3-11: Ce l’ho l’amici italiani.

Chapter Three

Pragmatic Markedness in L2 Italian

Fig. 3-12: Ce l’ho tempo.

Fig. 3-13: Ce l’ha undici anni.

77

78

Fig. 3-14: Ce l’ha sei (anni).

Fig. 3-15: Ce l’ho l’amici italiani.

Chapter Three

Pragmatic Markedness in L2 Italian

Fig. 3-16: Ce l’ho quattro cugini.

Fig. 3-17: Ce l’ho una zia.

79

80

Chapter Three

APPENDIX Speaker MK (L1: Tigrinya; DB “Progetto Pavia”) LTIIMK5

\Mk\ = io a:/arrivato a piedi in Kassalà dopo du/due o tre giorni così o vado Kartùm perchè eh ehm - io ce l’ho: la programma di: - vado, di viengo qua

LTIIMK6

Mk\ ++ eh - ho chiamato lui + “adesso - io ce l’ho: - un probl(i)ema - per andare Italia che cosa faccio adesso?”

LTIIMK8

\Mk\ Allora io non lo so ci sono/ ce l’ho il problema di/ delle parole

LTIIMK11

\Mk\ ce l’abbiamo due ghitare il basso e il normale \Mk\ le prove ce l’abbiamo una casa in via Kramer (xxx) sì eh noi troviamo qua \Mk\ sì di missionari ce l’abbiamo una casa piccola qua facciamo prova sempre^ \Mk\ sì stiamo provando anche il *reggae* inglis/inglese così stiamo provando perché eh: - eh noi no: andiamo solo al matrimonio, così anche ce l’abbiamo: feste del:/delle scuole (dell’Italia) per esempio ah: - eh - le scorse tre: settimane così siam/siamo andati a Treviglio \Mk\ eh quando sono arrivato in aeroporto ce l’avevo/l’avevo il numero telefonico della mia madre \Mk\ poi eh: ce l’abbiamo calcolo come si trova watt *watts*, kilowatt qualcosa^ \Mk\ un po’ di: ce l’abbiamo un po’ di lingua poi ce l’abbiamo un po’ di:, eh: eh= \MK\ no ce l’abbiamo il programma \MK\ anche - ce l’avevamo guerra tra – loro

LTIIMK12

\MK\ allora quando loro parlano ce l’hanno un vostro: lingua (bergamasco)

Pragmatic Markedness in L2 Italian

81

Speaker CH (L1: Chinese - wù dialect; DB “Progetto Pavia”) LCIICH6

\CH\ qua c’è un numero di terefono eh + eh + questo signore dire + dice eh c’è eh ce-l-hai eh (per) t(e)refono ah + ah + mhm ++ si/ signore (adesso) eh di eh sì ce/ eh eh qua c’è un un eh un terefono

Speaker XI (L1: Chinese- wù dialect; DB “Progetto Pavia”) LCIIX11

\XI\ io non ce l’ho coso da regalare a te

LCIIX12

\XI\ io ce l’ho foto + non è mi/ non è del genitori

LCIIX15

\XI\ io non ce l’ho/io ce l’ho nove anni +e poi n/ i altri non mi ricordo

LCIIX17

\XI\ io ce l’ho liblo però (questo)

CHAPTER FOUR “AND NEVERTHELESS THEY DO SPEAK ITALIAN”: PROSODIC EXPLORATIONS IN THE SPEECH OF DEAF IMMIGRANTS ELISA PELLEGRINO AND VALERIA CARUSO UNIVERSITY OF NAPLES “L’ORIENTALE”, ITALY

1. Introduction Starting from past research on the main features of deaf speech, our study focused on the prosodic competence of foreign deaf immigrants, taking into account both the degree of deafness and the general comprehensibility of their speech productions. We highlighted skills and strategies used to overcome their speaking impairment, and collected data on the deep influence that the pragmatic dimension of language may have even on deaf speakers, particularly during dialogical interaction.

2. Describing deaf speech: the existing medical research and perspectives for linguists There is a rich medical literature devoted to the analysis of deaf people’s speech production. These studies clarify both the physiologic aspects of mispronunciations and the input deficiencies responsible for the voice quality of these atypical speakers. The analyses carried out have indeed identified the basic components that characterize deaf people’s speech. Referring to the Prosody-Voice Screening Profile set down by Schriberg et al. (1990) for the evaluation of pathological speech, deaf speakers’ productions can also be described in terms of their vocal and prosodic features. In particular, Schriberg and colleagues describe voice in

84

Chapter Four

terms of loudness, pitch, laryngeal and resonance quality, while phrasing, rate, and stress belong to the prosodic component. Since Calvert’s (1962) investigation, deaf peoples’ voices have been described as “tense, flat, breathy”,1 while the adjectives “harsh” and “throaty” are usually used to refer to what Schriberg et al. (1990) have called the “laryngeal quality” of pathologic speech. Both tension and voice constriction are considered to be articulatory consequences of the kind of input these speakers rely on, which is tactile, rather than auditory. The breathy quality instead derives mostly from the “excessive force on plosives before a vowel”2 that deaf speakers usually produce, as we will see in the present study. The reverberation of sounds in the pharynx is, on its side, responsible for the resonance which produces articulations that are perceived as “hollow”,3 while researchers point out how bad nasalization is managed by these speakers: sometimes they fail to pronounce these sounds, and some others they hyper-articulate nasal sounds in contexts where these do not occur at all. Even more distinguishable are the prosodic traits of this atypical speech, whose main feature is the lack of fluency, described in terms of irregular phrasing, due to a general tendency to speak word by word without co-articulation, or sound blending.4 Fluency is also affected by the slow speech rate of deaf people, caused by prolonged vowels and long pauses between words.5 Stathopoulos et al. (1986) have highlighted the strategic use of pause duration for signaling sentence boundaries, which are marked by longer pauses than those used for words, instead of an adequate stress variation that should be used to control speech flow and mark the appropriate chunking. More generally, as Nickerson (1975) has pointed out, these speakers do not control stress contour, and tend to accent every single syllable they pronounce, a habit that is responsible for 1

Jessica M. Lenden and Peter Flipsen Jr, “Prosody and voice characteristics of children with cochlear implants”, Journal of Communication Disorders 40 (2007): 68. 2 Ibidem. 3 Daniel R. Boone, The voice and voice therapy (Englewood Cliffs, NJ: Prentice-Hall, 1971); Yehuda Finkelstein et al., “Peritonsillar abscess as a cause of transient velopharyngeal insufficiency”, Cleft Palate-Craniofacial Journal 30 (1993): 421–28. 4 Joanne Subtelny, Robert Whitehead and Nicholas A. Orlando, “Description and evaluation of an instructional program to improve speech and voice diagnosis of the hearing impaired”, The Volta Review 82 (1980): 85–95. Areti Okalidou and Katherine S. Harris, “A comparison of intergestural patterns in deaf and hearing adult speakers: Implications from an acoustic analysis of disyllables”, Journal of the Acoustical Society of America 106 (1999): 394–410. 5 Daniel R. Boone, “Modification of the voices of deaf children”, The Volta Review 68 (1966): 686–92.

Prosodic Explorations in Speech of Deaf Immigrants

85

the characteristic monotonous speech of many of the deaf. In addition, researchers6 have underlined how pitch control is influenced by the degree of the hearing loss itself. Bonetti et al. (2008) have analyzed prelingual mildly deaf adolescents, and found that they are able to control certain vocal features, such as speech fundamental frequency, intonation, accent, pauses and speech rate, proving that “prosody is not uniformly affected in the category of deafness, but that some prosodic features remain preserved with lower degrees of deafness.”7. Nonetheless, the bulk of medical studies is not concerned with real communication, and their analysis are carried out using word lists or small phrases read aloud as test material. This methodology actually allows easy comparisons and measurements for researchers that are mainly concerned with segmental aspects and the general intelligibility of speech. On the contrary, in our study we focus on the communicative skills and competence of a special group of deaf adults, namely deaf immigrants, who are in need of, and want to improve, their Italian in order to strengthen interaction with the hearing community of the new country they live in, and thus they ask for specific training not only in the written language but also in speech production. Some of them actually have hearing children, some others work in the public services, and some simply ask to be trained in vocal communication to accomplish the basic tasks of ordinary life. In this perspective, we have investigated the prosodic competence of three deaf immigrant women, and collected data on different aspects of communicative prosodic skills. While medical research can help those in need of health healing, and also in the field of hearing deficit, linguists should assist everybody asking for the improvement of their language skills. For this reason, we conducted a study analyzing the spoken productions of deaf foreign immigrants attending a course of spoken Italian at the University of Naples “L’Orientale” who had explicitly asked to be trained on small speech productions. The aim was to focus on communicative awareness and competence of those more impaired in vocal communication, and our results have proved the specific that pragmatics plays in their vocal linguistic capabilities. 6

Luka Bonetti, Vlatka Utoviü and Adinda Dulþiü, “Prosody in Speech Produced by Deaf Person”, Croatian Review of Rehabilitation Research 44 (2008): 1-13; Boone, “Modification of the voices of deaf children”, 686–92; Elaine Stathopoulos et al., “Intonation and pausing in deaf speech”, Folia Phoniatrica 38 (1986): 1–12; Nancy S. McGarr and Mary Joe Osberger, “Pitch deviancy and intelligibility of deaf speech”, Journal of Communication Disorders 11 (1978): 237–47. 7 Bonetti et al.,”Prosody in Speech Produced by Deaf Person”, 1.

86

Chapter Four

3. The study The present study was intended to analyse whether foreign deaf people have any prosodic competence in L2 Italian that enables them: x To produce a coherent phrasing of texts when reading aloud; and x To vary the pitch contour of different speech acts similarly to native speakers. In order to evaluate the role of the hearing loss level on the speech prosody of deaf immigrants, three deaf women who had previously attended an elementary course of vocal Italian were involved in the research: x Albina, 43 years old, with severe level of hearing impairment. She is a health-social worker who has been living in Italy for 11 years and has 11-year old hearing son; x Jurate and Reda, profoundly deaf women, 42-43 years old, Lithuanian, housewives, have been living in Italy for 12 years. The study also evaluated the speech productions of an Italian native speaker for a better estimation of the deaf performances. Initially we had administered to 25 graduated native Italians, attending a training course for teaching L2 Italian in order to ascertain Albina’s, Reda’s, and Jurate’s comprehensibility. The pre-test material consisted of a short of the reading by Albina, Reda, and Jurate extracted from the first test session used for the present research (see § 3.2.). The perception test had shown that Albina, who has the lowest level of hearing impairment and the greatest practice in spoken Italian, is the most comprehensible of the three. Albina’s oral production was rated sufficiently comprehensible by 80% of the Italian listeners, and scarcely comprehensible by the others. On the contrary, Jurate was considered scarcely comprehensible by 96% of the tested native speakers, and sufficiently by the other 4%. Only Reda was unanimously (100%) considered scarcely comprehensible. However, since we are not going to analyze the segmental features of deaf speech any longer, these results provide a quick estimation of this component.

3.1. The test The study was articulated into three phases. The first was devoted to exploring the possession of any communicative skill in reading a plain narrative text aloud. In this session we looked for evidence of any prosodic competence used for correctly conveying the meaning of a text, measuring pauses and searching for any coherent phrasing strategy. We used a 27word extract from an article of fashion, simplified for the deaf participants:

Prosodic Explorations in Speech of Deaf Immigrants

87

G-Star è uno stilista famoso e fa jeans molto belli. Oggi a Milano G-Star vende i suoi jeans a solo 20 euro, un prezzo molto basso. L’attrice Laura Chiatti fa pubblicità ai jeans di G-Star nel negozio di Milano. (G-Star is a famous stylist and produces beautiful jeans. Today in Milan he sells his jeans at only 20 euros, a very reasonable price. The actress Laura Chiatti promotes G-Star’s jeans in the shop in Milan).

The text was given for pre-reading, with images to help participants understand it. Then a multiple choice reading comprehension test followed. For the second task, participants were asked to pronounce three different speech acts, performing a set of utterances such as: x x x x x x

Che ora è? [What time is it?] (request); Esci subito! [Go out immediately!] (command); Puoi aprire la finestra? [Could you open the window?] (request); Ti ordino di aprire la finestra! [I command you to open the window!] (command); Apri la porta! [Open the door!] (command); Io vivo in Italia [I live in Italy] (assertion).

The aim of the task was to detect whether participants used different pitch contours in order to perform and differentiate between assertions, commands and requests. This kind of competence was also tested with another task consisting of semi-spontaneous conversation, in which deaf participants were asked to interact with hearing Italians never met before. The situation had elicited interaction and production of requests for the interlocutors. The questions produced in the semi-spontaneous speech production were then compared to those uttered in the acted speech task. For comparison, all the tasks were also performed by a native Italian.

3.2. Methods The L1 and L2 corpora were analyzed for single speech chains. For each chain we measured the number of syllables actually uttered, their duration, the lowest and highest F0 values, the occurrence of disfluencies, the length of silent pauses between the speech chains. On the basis of these measures, the following rhythmic-prosodic parameters were calculated: x

Articulation rate (AR), the ratio between the number of syllables actually uttered and phonation time (syll/s);

Chapter Four

88

x

x x

Speech rate (SR), the ratio between the number of syllables actually uttered and total time, including phonation time, empty or filled paused (syll/s). We considered empty pauses only silences that were longer than 100 ms; Fluency (F), the ratio between the number of syllables and the total number of speech chains; Tonal range in semitones.

Before examining the data from spectro-acoustic analysis, some preliminary methodological considerations are nevertheless needed. In our study, the corpus of speech produced by deaf participants was analyzed from the listener’s perception perspective, and not from that of the speaker’s production. In fact, if in speech by hearing people both the production and perception are strictly related,8 in deaf speakers the two perspectives do not always overlap (Fig. 4-1)

Fig. 4-1: Perceptual pauses vs. articulatory pauses.

What is perceived as silence does not necessarily correspond to physical pauses that are normally breathing or articulatory pauses.9 As in Fig. 4-1, the perceptual pauses (0.7-0.8.5 s) indeed may be the acoustic

8

Deng Li and Jianwu Dang, “Speech analysis: The production-perception perspective”, in Advances in Chinese Spoken Language Processing eds. Chin-hue Lee et al. (Singapore: World Scientific Publishing, 2007), 2-32. 9 Joost Schilperoord, It’s about time. Temporal aspects of Cognitive Processes in Text, Amsterdam-Atlanta, GA: Rodpi (Utrecht Studies in language and communication, 1996), 9.

Prosodic Explorations in Speech of Deaf Immigrants

89

correlates of compensating strategies necessary for preparing the following articulatory movements.

4. The results In this section, we will illustrate separately the results of the two tasks our subjects were involved in. Firstly we will focus on the data of the read speech task (see § 4.1.); then we will present the results of the task of acted speech (see § 4.2.).

4.1. The read speech task Data from the spectro-acoustic analysis of the read speech corpus show significant differences between the performance of the native speaker and those of the three deaf women. One of the prosodic features that seems to be most affected by the hearing impairment is the number of silences, which are much more frequent in the deaf, as is well known from existing literature. Jurate, Albina, and Reda pause respectively two, three and four times more often than the native Italian. (Elena 6 silences; Jurate 10; Albina 17; and Reda 24). The diverse use of silent pauses and the occurrence of disfluencies in the speech performed by deaf participants reveal significant differences among them in the composition of speech, in terms of fluency values, articulation, and speech rate. With only 6 silent pauses, lasting on average 0.450 s, the native speaker’s speech has the highest values of AR, SR, F and percentage of phonation time.

Fig. 4-2: Articulation Rate, Speech Rate and Fluency per speaker (E=Elena, A=Albina, J=Jurate, R=Reda).

Chapter Four

90

Phonation time Silence Disfluence

E

A

J

R

82,4%

72,8%

85,4%

76,6%

17,6%

27,2%

14,6%

22,9% 0,5%

Table 4-1: Composition of speech per speaker10. If we move on to consider the prosodic features within the group of deaf women, data from Fig. 4-2 show that there are parameters whose diverse values cannot be attributed to the degree of deafness (fluency), and features that are more affected by the level of hearing loss (articulation rate and tonal range). Likewise, the number of silences and variations in fluency do not correspond to the difference in the level of the hearing loss. Jurate and Reda, the two profoundly deaf women, are respectively the most and the least fluent speakers, though they are less comprehensible than Albina, as we have seen in the perception test in paragraph 3. On the contrary, Albina, the most comprehensible of the three deaf participants, with a severe level of hearing impairment, has an articulation rate about one syllable higher than the profoundly deaf participants, and also a wider tonal range (Albina 14,6 st; Jurate 10,4 st; Reda 9,2 st). In conclusion, the more profound the deficit, the longer the time to reach the articulatory targets, and the lower the tonal range. No relevant differences have been noticed instead in speech rate (about 2 syll./s). In order to evaluate the communicative skills in reading aloud a plain narrative text, we extended the analysis to the text phrasing strategy and compared the performance of the hearing speaker with those of our deaf participants. The chunking strategy used by the native speaker highlights the proper use of pauses, which reproduces the pragmatic organization of text information in topics and comments. All silent pauses last from around 300 to 600 milliseconds, and mark the boundaries between the theme introduced (e.g. G-star è uno stilista famoso) and what it is said about it (e.g. e fa jeans molto belli). The only remarkable exception is represented by a longer pause of ca 900 milliseconds at the end of the second paragraph (after un prezzo molto basso). This silence occurs when the

10

The mismatch between the number of silent pauses and the percentage of silence between the profoundly deaf Jurate and the hearing Elena depends on the fact that Jurate’s performance last longer (35 s) than Elena’s (13 s).

Prosodic Explorations in Speech of Deaf Immigrants

91

primary concept11 of the text, corresponding to the introduction of a new, not previously mentioned character, the actress Laura Chiatti. According to De Beaugrande and Dressler (1981) primary concepts are: “…points from which [text] accessing and processing can be strategically done”, since “concepts are […] steps in the construction of a continuity of sense”.12

The prolonged pause of 900 ms in the read speech of the native Italian signals a delay in the text processing, due to the radical shift in the conceptual architecture of the short text. Nr 1 2 3 4 5 6 7

Text chunks/Speech chains Silence (s) G-star è uno stilista famoso 0.308 e fa jeans molto belli. 0.573 Oggi a Milano G-star 0.410 vende i suoi jeans a soli 20 euro 0.438 un prezzo molto basso. 0.882 L’attrice Laura Chiatti 0.345 fa pubblicità ai jeans di G star nel negozio di Milano

Table 4-2: Text phrasing in the reading aloud task performed by the native speaker. On the contrary, the text read by Albina has an impressive number of pauses that present at least one overt marking strategy: the long silence introduced at the end of each paragraph. In this perspective, Albina seems unable to produce any other coherent phrasing, and her performance looks as if as she was aware only of the typographic aspect of the text.

11

Robert A. De Beaugrande and Wofgang U. Dressler, Introduction to Text Linguistics (London: Longman, 1981). 12 De Beaugrande and Dressler, Introduction to Text Linguistics, 95.

Chapter Four

92

Nr 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

Text chunks/Speech chains Silence (s) G-star 0.427 È 0.150 uno stilista 0.422 Famoso 0.346 e fa jeans 0.188 molto belli 1.202 Oggi a Milano 0.185 G star 0.627 vende i suoi jeans 0.820 A 0.180 solo 20 euro 0.840 un prezzo molto basso 1.292 L’attrice 0.153 Laura Chiatti 0.588 fa pubblicità 0.384 ai jeans 0.175 di G-star 0.771 nel negozio di Milano

Table 4-3: Text phrasing in the reading aloud task performed by Albina. However, by ordering the chunks according to pauses duration, a reading strategy is clearly recognizable. In fact, Albina tends to preserve syntactic boundaries, since shorter silences are used inside phrase units: [[[l’attrice]NP (0.153 s pause) Laura Chiatti]]NP (0.588 s pause) [fa pubblicità]VP (0.384 s pause) [[[ai jeans (0.175 s pause) [di G-Star]PP ]PP (0.771 s pause) [[nel negozio [di Milano]PP]PP]PP ]VP ]S. In this respect she meets the vocal identikit offered by the medical literature, according to which deaf speakers pronounce word by word without co-articulations and sound blends. However, the presence and length of the pauses indicate the use of a reading strategy.

Prosodic Explorations in Speech of Deaf Immigrants

93

Nr Text chunks/Speech chains Silence (s) 2 È 0.15 13 l’attrice

0.15

16 ai jeans

0.18

10 A

0.18

5 e fa jeans

0.19

7 Oggi a Milano

0.19

4 Famoso

0.346

15 fa pubblicità

0.384

3 uno stilista

0.422

1 G-star

0.427

14 Laura Chiatti

0.588

8 G star

0.627

5 di G-star

0.771

9 vende i suoi jeans

0.82

11 solo venti euro

0.84

6 molto belli

1.202

12 un prezzo molto basso

1.292

Table 4-4: Text phrasing by Albina: chunks ordered on the basis of their duration. Jurate’s reading has a more extended phrasing and significantly longer pauses, especially that of 940 milliseconds after the word euro. She tries to organize her reading in clear informative chunks, separating the argument—which functions as the main concept around which all other information is grouped—from its predicates. This becomes particularly evident in the second paragraph, after the word euro, when she produces the longest silence (940 ms) trying to make a reparation (repeating the verb vende), to preserve the text intelligibility, since she had interrupted her reading in many points before the interruptions being caused essentially by articulatory difficulties. Instead, in the first paragraph she had been almost successful in producing only two different chunks, a perfect separation between the topic (G-Star) and its comment (è uno

Chapter Four

94

stilista famoso e fa jeans molto belli). Unsurprisingly, the perfect division was interrupted only before the word belli, which starts with the sequence of a plosive plus a vowel, one of the crucial deficiencies in deaf people’s productions already highlighted in literature.13 Even if the third paragraph doesn’t show a similar clear phrasing, Jurate has proved to be aware of the pragmatic organization of texts and tries to respect it when reading aloud. Nr 1 2 3 4 5 6 7 8 9 10 11 12

Text chunks/Phonetic chains G-star è uno stilista famoso e fa jeans molto Belli Oggi a Milano G-star vende i suoi jeans a solo venti Euro vende prezzo molto basso L’attrice Laura Chiatti fa pubblicità ai jeans Di G-Star nel negozio di Milano

Silence (s) 0.155 0.255 0.921 0.391 0.405 0.288 0.940 0.752 0.678 0.236 0.226

Table 4-5: Text phrasing by Jurate. Such an awareness is not recognizable in the reading performed by Reda, who reads almost word by word, making very long pauses between paragraphs (1.117 s and 1.468 s), thus clearly revealing her articulatory difficulty.

13 Lenden and Flipsen, “Prosody and voice characteristics of children with cochlear implants”, 68.

Prosodic Explorations in Speech of Deaf Immigrants

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

N Text chunks/Speech chains G Star è uno Stilista Famoso E Fa Jeans Molto Belli Oggi a Milano G star vende i suoi jeans a solo 20 euro un prezzo molto basso L’attrice Laura Chiatti fa pubblicità ai jeans Di G star nel Negozio di Milano

95

Silence 0.207 0.422 0.285 0.295 0.25 0.385 0.175 0.158 1.117 0.393 0.450 0.263 0.524 0.153 0.417 0.85 1.468 0.450 0.542 0.292 0.150 0.45 0.167 0.15

Table 4-6: Text phrasing by Reda. Resorting to the most significant data collected and to the intelligibility rates from the pre-test, it is possible to notice that Albina, who has the lowest level of deafness, is significantly less fluent than Jurate. Both in reading for the pre-test extract14 and when she reads the whole article (see Fig. 4-2), Albina uses a phrasing strategy much more in line 14 Fluency data expressed in syllable/speech chain from the pre-test Albina 3; Jurate 6; Reda 2.

96

Chapter Four

with the existing literature findings on the slow, and scanned speech of deaf people. This speaking strategy is nevertheless successful, since Albina has proved to be the only “sufficiently” understandable of the three deaf during the pre-test, even if her phrasing is deeply affected by her general tendency to overuse pauses. On the contrary, the profoundly deaf Jurate, the most fluent deaf participant, adopting the same “pragmatic” phrasing strategy used by the native speaker, is nevertheless perceived as hard to understand by native Italians. Moreover, the articulation rate proves that Albina is the fastest deaf speaker, and consequently the most confident in producing segmental units, which have proved to be also sufficiently understandable to native speakers. On the contrary, Reda’s results show no significant communicative skills, either in phrasing or in fluency or in the intelligibility of what she says.

4.2. The acted speech task The spectro-acoustic data of the corpus of speech acts (1 assertion, 2 requests and 3 commands) confirms the results of the task of read speech. Once more, the prosodic features differentiating the speech of deaf women from that by the hearing participant are the greater number of silences, lower fluency, lower articulation and speech rate. If we consider the data concerning articulation rate and speech rate, it is possible to infer that the most significant difference between the two groups of participants are in the performance of the assertion and of the requests.

Fig. 4-3: Articulation rate per speaker and speech act.

Prosodic Explorations in Speech of Deaf Immigrants

97

Fig. 4-4: Speech rate per speaker and speech act.

Here, the hearing native speaker performs the two speech acts at a speech and articulation rate of about 6-7 syll./s, instead the values of these two parameters for deaf women are about 2-3 syll./s. The divergences decrease in commands where the gap between the hearing and the deaf participants is only 2 syllables long. As for the frequency of silences, the native speaker is always the most fluent, because she performs all speech acts as single phonetic chains.

Fig. 4-5: Fluency per speaker and speech act.

Chapter Four

98

If we compare the data within the group of deaf speakers, Reda is the most fluent. She segments only one request in two phonetic chains, and in doing so she preserves the integrity of the utterance unit of the speech acts. A completely different strategy is instead adopted by Albina, who segments all utterances in one or more phonetic chains. Reda’s higher fluency shows that she is more similar to the native speaker than Albina, from a prosodic point of view. Nevertheless her performance is characterized by segmental misarticulations and disfluencies that heavily affect the message intelligibility and its communicative efficacy. Similarly to the results of the reading task, there are no meaningful differences in speech rate. The severity of deafness seems instead to have an impact on the articulation rate since, with the exception of only one command (Apri la porta), Albina, the severely deaf one, is always the fastest speaker. It is important to underline, however, that the lower values of articulation rate by profound deaf women do not correspond to more accurate articulatory movements, as usually happens in the speech by hearing speakers,15 but they are mainly due to their difficulty in producing the required phones. As for the intonational structure employed to express the three speech acts, the native speaker differentiates very precisely her pitch contours on the basis of the communicative intent to convey: x x x

An assertion has an initial pitch peak with a gradual lowering on the remainder of the utterance; A request has an initial pitch peak with final rising contour; A command has an initial pitch peak with final falling contour.

In every case the intonation unit is the utterance. The deaf women use pragmatic and prosodic strategies that do not correspond to those used by Italian native speakers. Albina, for example phrases the utterances in words, and Jurate is not able to vary her pitch contour to convey the three specific communicative intents. The most extreme case is represented by Reda, who pronounces all speech acts with a flat pitch contour. She has also the lowest tonal range.

15 Massimo Pettorino and Antonella Giannini, “Il parlato dei mass media”, in La Comunicazione Parlata 3, eds Massimo Pettorino, Antonella Giannini and, Francesca. M. Dovetto, (Napoli, OPAR, Università degli studi di Napoli l’Orientale, 2010), 71-83.

Prosodic Explorations in Speech of Deaf Immigrants

Assertion Request Command

E 13.1 12.6 12.7

A 16 18.75 8.9

J 10.4 7.84 9.2

99

R 6.1 5.4 5.3

Table 4-7: Tonal range per speaker. Concluding, in order to collect exhaustive data, we gave two sentences of different length for command and request. In this way it was possible to notice the lack of significant variation in articulation and speech rates of deaf speakers’ productions with sentences of different syllable length expressing the same speech act, such as the commands: Apri la porta and Ti ordino di aprire la finestra. For example the speech rate of Apri la porta is 2,4 syll./s and Ti ordino di aprire la finestra 3,1 syll./s in the performance of Albina, while in Jurate’s they are respectively 2,4 and 2,1 syll./s. The same does not hold for the native speaker, who varies her rate in order to perform the sentences as single utterances, and coherently realizes single pragmatic units. There are significant exceptions to this rule though, since in at least one case Reda is almost as fast as the native speaker, both in speech and articulation rate, performing the command Apri la porta at a speech and articulation rate of 4,5 syll./s. And Jurate, surprisingly, is also as fluent as the native Italian performing Esci subito and Che ora è?. The only performances showing some prosodic competence are indeed those produced by the least intelligible participants of the study, a fact that compels us to draw the conclusion that the scanning strategy used by Albina is the only one capable of overcoming the segmental impairments these speakers have, and which allows her to perform sufficiently understandable speeches. But the segmental strain affects prosody, which is incorrectly managed. On the contrary, the good performances shown by the profoundly deaf prove the existence of a prosodic competence, independent of their real production skills and level of deafness. This competence is linked to the implicit awareness that communication develops through units, and unity is given in different ways depending on the message channel, such as prosody, among many others, for spoken communication. In the task of acted speech, the acoustic deficit, which prevents deaf participants from receiving sound feedback, is responsible for their rhythmic irregularity and lack of control on the pitch contour. These results are partially overturned in the task of semi-spontaneous interaction,

100

Chapter Four

during which all of the three deaf speakers are able to produce a request using the right prosodic contour: an initial pitch peak with final rising contour (Figures 4-6, 4-7 and 4-8).

Fig. 4-6: Pitch contours in Albina’s request during the semi-spontaneous speech production. Mi chiamo Albina e tu? (“My name is Albina, what about you?”).

Fig. 4-7: Pitch contours in Jurate’s request during the semi-spontaneous speech production. Come stai? (“How are you?).

Prosodic Explorations in Speech of Deaf Immigrants

101

Fig. 4-8: Pitch contours in Reda’request during the semi-spontaneous speech production. Come stai? (“How are you?”).

5. Conclusions The spectro-acoustic analysis of read and acted speech corpora show significant differences between the performance of the hearing native speaker and of those of the 3 deaf women. In both tasks, the native speaker is always the most fluent, and fastest, in terms of articulation and speech rate. As for the differences within the group of deaf women, the data on read speech show that there are parameters whose diverse values cannot be attributed to the degree of deafness (fluency), and features that are more affected by the level of hearing loss (articulation rate and tonal range). The more profound the deficit, the narrower the tonal range, and the longer the time to reach the articulatory targets. Nevertheless, the lower values in the articulation rate are mainly due to the deaf people’s difficulty in vocal production, but the nearly native-like fluency in Reda’s acted speech task—the least comprehensible deaf person, unable even to produce a coherent text phrasing—seems to prove the existence of a communicative awareness and competence detached from the articulatory difficulties that prevent efficient spoken communication. While the acoustic deficit deeply compromises the segmental speech productions, the acted speech task highlighted some prosodic competence (fluency) that preserves communicative units, and in the semi-spontaneous interaction even the most profoundly deaf people produce the required prosodic curves for questions. Pragmatic awareness is also clear in the text phrasing of a profoundly deaf person.

102

Chapter Four

The study provides evidence that the search for intelligibility and communication of contents in deaf speakers reduces prosodic control, but prosody has nevertheless proved to be a linguistic skill that these speakers do possess. Moreover, we have focused on communicative ability and searching for real speech behaviours, and not for artificial lab performances, as Lenden and Flipsen (2007) have correctly declared, asking for a new kind of research into the speech productions of deaf speakers.

CHAPTER FIVE ABOUT ‘ETHNICS’ ITALIAN’S PROSODIES: DATA AND OBSERVATIONS DOMENICO RUSSO UNIVERSITY OF CHIETI – PESCARA, ITALY

1. The study of IRA The content of this paper originates in an analysis of “Italiano regionale abruzzese” (IRA) (Russo 2004). The aim of the analysis was to establish the distinctive features of IRA, as has been done in the case of the Florentine and Paduan languages (Magno Caldognetto et al., 1978), and also for the Neapolitan language (Maturi 1988). To discover the distinctive features of IRA it was important to define what “regional Italian”1 really is. The concept of “Italiano regionale”, has in fact, been present in all studies of Italian linguistics, but there has always been doubt over its meaning. There are only two levels of linguistic analysis which define it: the semantic2 and prosodic levels. The latter level, which is the focus of this study, has been regarded as the most

1

Cf. Alberto Sobrero, “Italienisch: Regionale Varianten”, in Lexicon der Romanistischen Linguistik (LRL), eds. G. Holtus et al. (Tübingen: Niemeyer, 1988), 732-48 and Tullio Telmon, Guida allo studio degli italiani regionali (Alessandria: Edizioni dell’Orso, 1990), Domenico Russo, “I ‘colori’ della semantica regionale. Indagine sui termini del lessico alimentario abruzzese,” Plurilinguismo. Contatti di lingue e culture 8 (2001): 111-26 and Russo, “Per lo studio dei lessici degli italiani regionali. I Materiali per un lessico alimentario abruzzese”, in Studi e ricerche di terminologia alimentaria, ed. C. Consani (Alessandria: Edizioni dell’Orso, 2001), 27-51 and 91-410. 2 On the semantic level cf. Russo “I ‘colori’ della semantica” and Russo “Per lo studio.”

104

Chapter Five

important spoken aspect of regional Italian.3 But in this case there is also a lack of empirical data pertaining to it.4 In our study of IRA we examine, in particular, the data concerning the starting point, the acme values, the pitch trend and its number, as well as the terminal contour and average utterance duration. We are in a position to establish differences between the regional Italian of Naples and of Abruzzo, as southern varieties of Italian, in comparison with northern regional Italian.

2. The ‘New Italian Voices’ corpus In the last few years, in Abruzzo as well as the other regions of Italy, a new and important variety of Italian has been added which we call ethnic Italian, the variety brought to the country by immigrants. Additionally in this new kind of language, which is a galaxy of super-diversity (Vedovelli 2009), prosody is the principal distinguishing feature. All the reasons which lead us to study the prosody of regional Italian — for example, the importance of regional varieties regarding standard Italian, fields of research on processes of language acquisition, sets of prosodic data, important components in studies of the voice in Italian culture (De Dominicis 2002) — remain the same for studies of ethnic Italian. With the kind collaboration of the educational institutions and cultural associations of Abruzzo we have gathered a corpus of interviews with foreign native speakers (New Italian Voices). The interviews were divided into two parts: the first part was a free conversation between the interviewer and the subjects about their relationships with Italy and the Italian language; and the second part consisted of ten utterances, spoken three times each, according to our protocol.5 3

Cf. Sobrero “Italienisch”, 734; Tullio Telmon, “Gli italiani regionali contemporanei”, in Storia della lingua italiana, Vol. III, Le altre lingue, eds. Luca Serianni and Pietro Trifone (Torino: Einaudi, 1994), 613-14; Silvana Ferreri, “Italiano regionale di Sicilia: dove si parla e dove se ne parla”, Italiano e Oltre X, 4 (1995): 44-45. As for Abruzzi, Paolo D’Achille, “Attraverso i ‘ponti’ dell’Abruzzo e del Molise”, Italiano e Oltre XI, 5 (1996): 285-91. 4 See the important work of Pier Marco Bertinetto, Strutture prosodiche della lingua italiana. Accento, quantità, sillaba, giuntura, fondamenti metrici (Firenze: Accedemia della Crusca, 1981), and Federico Albano Leoni, “Sulla voce”, in La voce come bene culturale, ed. Amedeo De Dominicis (Roma: Carocci, 2002), 3962. 5 We remind you that the referred protocol foresees the elicitation of ten sentences (three declarative: Giovanni arriva; Andiamo al cinema; Giovanni mangia la mela;

About “Ethnics” Italian’s Prosodies: Data and Observations

105

The first and most important result that emerged from our recordings was the fact that no correlation is possible with Italian data, neither Abruzzese, nor of any other region. There is a great lack of common constitutive conditions between the utterances as shown in the corpus and those of standard or regional Italian.

3. From utterance to voice Based on this evidence we felt it necessary to change all our methodological criteria of analysis. We noted several things: for example, when an Italian utterance is analysed all the constitutive conditions of that utterance are neutralised therefore making the prosodic element clear and manageable. On the other hand, with the utterance of a non-native speaker this phenomenon does not occur because all the constitutive conditions are active and clearly visible. Any abstraction regarding the data is not only impossible, but also not particularly significant. For instance, in the analysis of two native speakers, the vocal characteristics are neutralised, whereas in the analysis between a native and non-native speaker the characteristics are not neutralised.6 As has already been established, the vocal characteristics are almost never able to be associated with the discrete and recursive syntagmatic element.7 The vocal characteristics are almost always global. Even when we can identify them with some discrete elements (such as Russian or Chinese phonemes), they are traces of linguistic contact which have deep repercussions on the prosody of the utterances. Another consideration is that when we compare two utterances of two native speakers, the meanings of the vocal characteristics are neutralised. On the contrary, particular vocal characteristics of a non-native speaker do not carry the same meanings as for a native speaker. And so once again it is necessary to establish the consistency of these components in the whole prosodic configuration of the utterances. A third ‘hidden’, but most interesting, piece of evidence is the neutralisation of the editing operations. In a standard interlinguistic comparison the value of the editing operation is near to zero. This is three polar interrogatives: Giovanni arriva?; Andiamo al cinema?; Giovanni mangia la mela?; and four interrogatives wh-: Chi arriva?; Dove andiamo?; Chi mangia la mela?; Cosa mangia Giovanni? 6 In particular, we refer to Ivan Fonagy, La vive voix. Essai de psycho-phonétique (Paris: Payot, 1983) and to Albano Leoni “Sulla voce”. 7 Cf. De Dominicis, “La voce come bene”. For a general theoretical perspective, refer to Fonagy, La vive voix.

106

Chapter Five

absolutely not the case with ethnic utterances. An ethnic utterance which undergoes analysis by a linguist is automatically rewritten from the ‘beastly’ to the ‘beautiful’. Everything that emerges from the rewriting is taken as a constitutive element of the utterance, whether positively or negatively.8 Finally, to analyse the ethnic utterances means making the set of their constitutive conditions explicit. Allied to this, we encounter all the problems tied to the Saussurian actes de parole as well as the relative ‘youth’ of the instrumental phonetic applied to spontaneous speech. However, as Aristotle arguing against the Sceptics taught us, comparison between utterances is always possible, especially today, thanks to physical, mathematical and technologically refined models, in order to study spontaneous speech. Now, more than in the past, we are in a position to clearly establish whether two linguistic elements are equal or similar or only familiar. Therefore, when we study one ethnic utterance we have to be able to explicitly say, for example: a) which vocal characteristics are organic in nature (a Lithuanian woman is different from a Chinese woman, and both are different from a Romanian woman); b) which vocal characteristics are acquired through mother-tongue (native language or dialect, socio-cultural environment, process of instruction); c) which vocal characteristics are acquired due to contact (the differences between a Rome-based Romanian and a Romanian who lives in Pescara). Another issue to resolve is to establish the degree of auto-editing activated by the speaker during the recording session. The phenomenon of ‘unconscious camouflage’ of the speaker when he/she is asked to pronounce an utterance in a foreign language is well known. Briefly, it is always necessary to discern how much is to be ascribed to dynamic variations which are inter-spoken, and how much ascribed to those variations which are intra-spoken.

4. Some methodological choices and the problem of the ‘model’ For us, the best methodological path is a parametric method.9 In this method we measure the number of utterances, their duration, the number 8

The point is very important cf. Federico Albano Leoni, Dei suoni e dei sensi. Il volto fonico delle parole (Roma: Carocci, 2009). 9 Cf. Andrea Paoloni, “La voce come elemento di identificazione della persona”, in La voce come bene culturale, ed. Amedeo De Dominicis (Roma: Carocci, 2002), 125-39.

About “Ethnics” Italian’s Prosodies: Data and Observations

107

of empty pauses and their average duration, as well as the frequencies of the vowel formants and their duration. The essential point in experimental phonetics is therefore the comparison between two utterances. The standard method is to use a model to invoke comparison, but we must take into account that a model is always something constructed, particularly in the case of the Italian language in which the studies of regional Italian language show that a standard model does not exist. In other words, we must accept the fact that when a model is used in experimental phonetics this model is never ‘real’. The reality is that the linguist chooses a particular linguistic variety and considers this as a ‘real’ model. This means that the results of our comparisons do not correlate with “lingua viva e vera” as Nencioni stated, but with the coordinates that we have used to establish our model. The scientific method requires protocols which must be repeated and verified, so the coordinates of the construction of the model must be explicit. In the case of ethnic utterances, the situation becomes even more complicated, because there are two models. On the one hand there is the mother tongue model, whilst on the other there is the target language model. It is clear that the construction of the first model causes the same problems as the construction of the second. Even if all these difficulties are overcome, the result remains a modest one. In experimental phonetics, from an epistemological perspective, the way of comparing the data following the method many to one, where the one is a model, is not productive. It must be recognised that the most useful comparative method is many to many. We can observe in the comparison of L1 utterances that the many to many method has been found to be feasible.10 On the contrary, regarding comparisons between Italian L1 and L2 utterances, the same method is very difficult today, above all because of the lack of sufficient data. Faced with this difficulty, the best solution seems to be to imitate the process of the birth of languages, which develop according to the configuration that mathematicians call pergola. In other words it is necessary to start from a few simple pieces of data and then gradually build a configuration which is both complex and expansive.

10 Cf. Massimo Pettorino, “I cambiamenti della lingua italiana,” in La voce come bene culturale, ed. Amedeo De Dominicis (Roma: Carocci, 2002), 141-57.

108

Chapter Five

5. Can a holistic theory of speech be of help? Two important arguments support this hypothesis. The first is the fact that linguistics has increasingly concerned itself with spontaneous speech, and in particular with the prosodic and intonational components. The second is that prosody is now regarded less as a suprasegmental phonological feature than as a connective component of the elements of an utterance (i.e. turn-taking, syntactic disposition, lexical form, phonic form, etc.).11 This method certainly seems to ask more questions than it in fact answers because it means adopting a ‘star’-type analysis in which the nucleus — the prosody — changes position continually. For this reason it becomes more interesting to those who support the idea of a nonsegmental, holistic, physiognomical speech model.12 The description of the phonic images of the ethnic utterances appears very significant. We find ourselves in favourable circumstances, in the presence of totally new images as opposed to those already noted. There is no doubt that faced with ethnic utterances “we distinguish between two faces even if they are unknown”,13 but above all “the more contextual elements we have available the easier it will be to recognise these faces”.14 The critical point of the whole argument is in the information the context offers on the intelligibility of the data. The analysis of ethnic utterances once again proves what experiments in progressive decontextualisation and de-cotextualisation have already shown.15 In the case of spontaneous ethnic speech, the levels of discernibility of the sequences can be extraordinarily low; but the case of semispontaneous speech — sequences which are in any case intelligible — can be even more interesting. The contextual elements that determine the form of these types of sequence are ‘mute’ to the ear of the analyst. With an ethnic utterance the only thing we can do is to recognise it as such; if we listen to it for the first time in the language laboratory it is difficult to ascertain, through the vocal characteristics, the ‘basic facts’ such as sex, age and nationality, with sufficient certainty. Some points of holistic theory will be of great help to our work. First of all, there is the notion of the ‘speech-act’ (Fr. enoncè; it. 11

Cf. Amedeo De Dominicis, Intonazione. Una teoria della costituenza delle unità intonative (Roma: Carocci, 2010). 12 Cf. Albano Leoni, Dei suoni e dei sensi. 13 Albano Leoni, Dei suoni e dei sensi, 166. 14 Ibidem. 15 Cf. the review given by Albano Leoni, Dei suoni e dei sensi, 175-6 and n. 23.

About “Ethnics” Italian’s Prosodies: Data and Observations

109

enunciato; which other authors call text, phrase, utterance, proposition, syntagma, phonological word, word, etc.).16 The best hypothesis for us is that any speech-act is a set of modifications of a physical state and that all our macro- and/or micro-linguistic hypotheses are the expressions of flow invariance. Secondly, there is the notion of context and, in spontaneous speech, the notion of cotext. It is in context and cotext where we find what is more than the sum of the parts of the utterance. Thirdly, a refined representation of the linguistic sequences depends significantly on the results of self-checking by the speakers on their production and their comprehension. We know that the speaker is his or her own first interlocutor. Fourthly, the results of the dialogic adjustments of both speakers in spontaneous speech must not be forgotten. Many editing operations are always in action, especially in the case of ethnic utterances and L2 utterances. Lastly, and briefly, there is the hypothesis which claims that “a phonic image, different from all other phonic images, or equal to some other, or itself, in different conditions, expresses its diversity or its identity through its property which is the voice, the syllable and the prosody, consubstantial with it”.17

6. The importance of the differences The general conditions of our analysis should be those as quoted by Georges Boulakia18 according to the fourth chapter of de Saussure on linguistic value: “The important thing in the word is not the sound alone but the phonic differences that make it possible to distinguish this word from all others, for differences carry signification”,19 and that following “it is impossible for sound alone, a material element, to belong to language. It 16

Concordance between linguists on this point is not unanimous. We limit ourselves here to the utility of a notion of an utterance close to the one proposed illo tempore, but good ideas are eternal, by Sergej Karcevskij, “Sur la Phonologie de la Phrase.” Travaux du Cercle Linguistique de Prague IV (1931): 188-227 and opportunely resumed by Domenico Di Russo, Intonazione e modalità nella significazione. PhD diss., Università di Roma “La Sapienza”, 2011. 17 Albano Leoni, Dei suoni e dei sensi, 183, my italic. 18 Georges Boulakia, “Linguistica e fonetica: senza voce o mezza voce?,” in La voce come bene culturale, ed. Amedeo De Dominicis (Roma: Carocci, 2002), 63. 19 Ferdinand de Saussure, Course in General Linguistics, trans. Wade Baskin (New York: Philosophical Library, 1959), 117.

110

Chapter Five

is only a secondary thing, substance to be put to use”,20 and also that the signifier is something incorporeal, constructed by differences that distinguish their image from the other. In a language in fact, as Boulakia cites again, “there are only differences” (italics my own) to which we must also add “Even more important: a difference generally implies positive terms between which the difference is set up; but in language there are only differences without positive terms”21 (italics my own) because as is well-known only “when we consider the sign in its totality, we have something that is positive in its own class”.22

7. Conclusions Returning to our corpus of phonic ethnic ‘images’, there are no sufficient data to permit any general observation. Our corpus therefore allows a series of methodological observations. The most important observation for us consists in noticing that the research into invariance that constructs a ‘phonic face’ is impossible without comparison between utterances whose constitutive coordinates are explicit, or at any rate are able to be explicit. For example, if we have two Albanian speakers who utter sequences that we recognise as non-regional Italian ethnic utterances (because their vocal characteristics are neutralised, they have the same semantic content and they are virtually the same from a level of auditory perception), only instrumental analysis is able to catch the nuances of difference, but also only the knowledge of context (in this case, the differences in the biographical data such as age, profession, period of residence in Italy) allows us to attribute these differences to issues of editing (this is the case of Alba 1 versus Alba 2 in the utterance Giovanni arriva, cf. Tab 1). In the same way, only the knowledge of context permits us to assign the form of an utterance to the effect of persistency of the mother tongue (as in the case of Giovanni mangia la mela, uttered by Alba 2, versus the more ‘Italian’ utterance of Alba 1). These are only two examples of characteristics from the utterances in our corpus, established through the differences in the whole of the utterance and through every element of the context. The point is that not only words or syntactic schema, but also that everything in an utterance is language. The difference of a few Hz in a 20

Ibidem. Ivi: 120. 22 Ibidem. 21

About “Ethnics” Italian’s Prosodies: Data and Observations

111

starting point or a few milliseconds in the duration of two or more sequences can reveal themselves, in the light of contextual conditions, as clear traces of elaborate operations of linguistic construction and control.

Fig. 5–1: Giovanni arriva.

Fig. 5–2: Giovanni mangia la mela.

PART II THE ROLE OF L1

CHAPTER SIX GEMINATE TIMING IN THE SPEECH OF ESTONIAN L2 LEARNERS OF ITALIAN CHIARA CELATA SCUOLA NORMALE SUPERIORE PISA, ITALY

AND LIDIA COSTAMAGNA UNIVERSITY FOR FOREIGNERS PERUGIA, ITALY

1. Introduction The research project described here focuses on the production of geminates in the speech of native Estonian learners of L2 Italian.1 The

The authors are grateful to Na Zhi, Irene Ricci and Chiara Bertini for their collaboration in the digitalization and annotation of the corpus. 1 The research is funded by the University for Foreigners Perugia (ex 60% 20102012 on behalf of L. Costamagna) within the project “Acquisizione della fonologia dell’italiano come L2: la geminazione consonantica” / “Acquisition of L2 Italian phonology: consonant gemination”. Gemination in the learners’ variety is among the most important research themes at the University for Foreigners Perugia, where a long standing research project on the acquisition of L2 Italian phonology is being developed. See e.g. Lidia Costamagna and Stefania Giannini, eds., La fonologia dell’interlingua: principi e metodi di analisi (Milano: Franco Angeli, 2003); Lidia Costamagna and Giovanna Marotta, eds., Acquisizione linguistica e categorie fonologiche (Pisa: Pacini, 2008); Lidia Costamagna and Stefania Scaglione, eds., Italiano: Acquisizione e perdita (Milano: Franco Angeli, 2008). The crosslinguistic conditions investigated over the past few years vary for the involved L1s (Romance, Germanic), the processing modality (production, perception) and the acquisitional setting (guided and non-guided acquisition, attrition under sociolinguistic pressure, e.g. Stefania Giannini and Lidia Costamagna, “Acquisizione di categorie fonologiche e diffusione lessicale del mutamento linguistic: affinità strutturali”, Archivio Glottologico Italiano 83 (1998): 150-87; Chiara Celata and Jessica Cancila, “Erosione sociolinguistica nell’italiano di emigrati. Il caso della lunghezza

116

Chapter Six

purpose of the project is to analyze prosodic tuning during the acquisition of the Italian singleton-geminate consonant contrast.2 The work is still in progress; this contribution presents and discusses some preliminary results. As consonant gemination is present in Estonian (though with different characteristics in comparison with Italian), the research does not deal with the introduction of a totally new feature in the interlanguage of Estonian learners; on the contrary, it focuses on the process of restructuring native prosodic patterns according to the target language patterns.

consonantica”, in Italiano: Acquisizione e perdita, eds. Stefania Scaglione and Lidia Costamagna (Milano: Franco Angeli, 2008), 165-87; Chiara Celata and Jessica Cancila, “Phonological attrition and the perception of geminate consonants in the Lucchese community of San Francisco (CA)”, International Journal of Bilingualism 14/2 (2010): 1-25. Length has also been investigated with respect to the acquisition of other consonant features such as affrication. See e.g. Chiara Celata, Acquisizione e mutamento di categorie fonologiche: le affricate in italiano (Milano: Franco Angeli, 2004); Chiara Celata, “Allophonic variation in speech perception”, in Recent Research in Second Language Phonetics/Phonology: Perception and Production, eds. Michael A. Watkins andreia S. Rauber and O. Baptista (Newcastle-upon-Tyne: Cambridge Scholars Publishing, 2009), 64-80; Lidia Costamagna, “The acquisition of Italian L2 affricates: the case of a Brazilian learner”, in New Sounds 2007: Proceedings of the Fifth International Symposium on the Acquisition of Second Language Speech, eds. Andreia S. Rauber, Michael A. Watkins and Barbara O. Baptista (Florianópolis: Federal University of Santa Catarina, 2007), http://www.nupffale.ufsc.br/newsounds/proceedings, ISBN: 97885-98703-04-6; Lidia Costamagna and Cristina Montilli, “Le affricate italiane: percorsi di acquisizione”, in Acquisizione linguistica e categorie fonologiche, eds. Lidia Costamagna and Giovanna Marotta (Pisa: Pacini, 2008), 67-92. 2 A distinction is sometimes made between consonant “length” and “gemination”, depending on whether reference is made to the monophonematicity or biphonematicity, respectively, of long consonants in Italian. See e.g. Zࡅ arko Muljaþiü, Fonologia della lingua italiana (Bologna: Il Mulino, 1972); Michele Loporcaro, “On the analysis of geminates in Standard Italian and Italian dialects”, in Natural Phonology: The State of the Art, eds. Bernhard Hurch and Richard A. Rhodes (Berlin: Mouton de Gruyter, 1992), 153-87. The phonological interpretation of Italian geminates is still a matter of debate. See e.g. Claudio Zmarich, et al., “Consonanti scempie e geminate in Italiano: studio acustico e cinematico dell’articolazione linguale e bilabiale”, in Atti del III Convegno Nazionale dell’Associazione Italiana di Scienze della Voce (AISV), eds. Veronica Giordani, Valentina Bruseghini and Piero Cosi (Torriana: EDK Editore, 2007), 151-63. In this contribution, the two terms (“long” and “geminate”, as well as their counterparts “short” and “single”) will be used without specific reference to any of the two theoretical positions.

Geminate Timing in the Speech of Estonian L2 Learners of Italian

117

We have been inspired by the model of geminate timing in prosodically different languages proposed by Ham (2001). According to Ham’s (2001) proposal, distinct phonological and phonetic factors explain both recurrent and idiosyncratic aspects of quantity distribution in different languages. Among the phonological factors are the characteristics of the general prosodic hierarchy (i.e. languages may rely on either the syllable, mora, or foot) and the restrictions on segment order in the VC sequence (i.e. languages may or may not allow contrastively long vowels to surface before geminate consonants). From a phonetic point of view, language-specific factors (such as allophonic vowel shortening in closed syllables) mix with hypothetically “universal” factors, such as those predominantly determined by the physical restrictions of the vocal apparatus (e.g. the tendency to shorten stressed vowels followed by final unvoiced plosives, compared to when they are followed by voiced plosives). All of these factors interact with one another; the domain of quantity is probably one of the most informative with respect to the phonetics/phonology interface. The cross-linguistic perspective (both in the sense of the comparison among different quantity systems, and of the analysis of learners’ interlanguage) may be useful to shed light on some of these aspects. The analysis presented here has been carried out exclusively on timing relations in consonant gemination (i.e. duration variations). The focus of this contribution is on the temporal relationships between adjacent consonant and vowel segments as they are deducible on an acousticspectrographic basis; the existing literature on the articulatory and cinematic characteristics of gemination (as investigated in electropalatographic, radiographic, electromyographic and electromagnetic research) remains in the background3. The corpus of acquisitional data analysed in this contribution is very limited in scope; the segmentation, annotation and analysis of a larger database is still in progress. This contribution illustrates the general

3

One study worth mentioning with respect to Estonian is Ilse Lehiste, Katherine Morton and Mark A.A. Tatham, “An instrumental study of consonant germination”, Journal of Phonetics 3 (1973): 131-48, which proposed a pioneering electromyographic description of quantity contrasts. On the other hand, Italian is currently the focus of a noteworthy and very productive line of research on the acoustic and cinematic characteristics of geminate consonants. For a recent contribution, see Claudio Zmarich, et al., “Speech timing organization for the phonological length contrast in Italian consonants”, in Proceedings of Interspeech 2011 (Firenze, August 28-31, 2011), 401-04.

118

Chapter Six

working hypotheses, the nature of the data, and a first experimental approach to gemination in the interlanguage of the Estonian learners.

2. Theoretical and empirical premises 2.1. Consonant quantity in Estonian and Italian In Estonian, all 9 vowels and 14 consonants, with the exception of /v/, can occur in three phonetically distinct quantities.4 In common use, the three quantitites are referred to as short, long and extra-long consonants or vowels (Q1, Q2 and Q3 abbreviations are also used).5 The presence of a ternary length opposition in Estonian and other languages is considered a clue to the existence of trimoraic (or “super-heavy”) syllables.6 Not all quantity contrasts may be achieved in all phonotactic contexts. In Estonian, the first stressed syllable of polysyllabic words is the context where the maximum number of oppositions may be realized.7 A typical example of a minimal pair in Estonian is the following (1), where the alternations may concern both the vowel (1a) and the consonant (1b): (1) a. koli “rubbish” – kooli “(of) school” gen. sg.– koooli “(of) school” part. sg. 4

Ilse Lehiste, Consonant quantity and phonological units in Estonian. (Bloomington: Indiana University Publications, 1966); Olle Engstrand and Diana Krull, “Durational correlates to quantity in Swedish, Finnish and Estonian: Crosslanguage evidence for a theory of adaptive dispersion”, in Festschrift on the Occasion of Björn Lindblom’s 60th Birthday, eds. Olle Engstrand and Klaus Kohler, Special issue of Phonetica, 51 (1994): 80-91; Ilse Lehiste, “The search for phonetic correlates in Estonian prosody”, in Estonian Prosody: Proceedings of the International Symposium on Estonian Prosody, eds. Ilse Lehiste and John Ross, (Tallinn: Institute of Estonian Language, 1997), 11-33; Martin Ehala, “Estonian quantity: Implications for Moraic Theory”, in Generative Approaches to Finnic and Saami Linguistics, eds. Diane Nelson and Satu Manninen (Stanford: CSLI, 2003), 51-80. 5 In the phonemic inventory of Estonian there is only one series of voiceless, unaspirated plosives, with three contrasting lengths; orthographically they are usually represented as (short consonant series),

(long consonant series) and (extralong consonant series). Estonian lacks affricate phonemes. 6 Bruce Hayes, “Compensatory lengthening in moraic phonology”, Linguistic Inquiry 20 (1989): 253-306. 7 Lexical stress in Estonian is always on the first syllable, with rare and lexically determined exceptions, and with the exclusion of most loan words.

Geminate Timing in the Speech of Estonian L2 Learners of Italian

119

b. koli “rubbish” – kolli “(of) bugaboo” gen. sg. – kollli “(of) bugaboo” part. sg. In intervocalic position between the first and second syllable, long or extra-long consonants preceded by long vowels (2a) or extra-long vowels (2b) are allowed: (2) a. saatta “send” inf.: sVVCCCa8 b. hoooppis “entirely”: hVVVCCCis9 In sum, if compared with similar languages such as Finnish and Japanese, Estonian features a phonological distinction among three grades of segmental quantity in some phonotactic contexts; like several other languages with a tendency toward moraic isochrony, Estonian shows quantity distinctions on both vowels and consonants and therefore allows for the sequential combination of long vowels and consonants.10 However, in the phonetic implementation there are strong compensatory oscillations which make the distinctions much more uncertain and variable than would be expected. For example, in reporting on the case of the word hoooppis (see above, 2b), Lehiste herself notes that, in her reference corpus (see below, § 4.1. for details), the bilabial plosive in Q3 has a duration of 100 ms, which is comparable to the duration of most corresponding plosives in Q2; similarly, the preceding vowel, measuring 110 ms, is significantly shorter than a corresponding vowel in Q3 when followed by short C, such as the vowel in koooli, which has a duration of 260 ms. Another significant example of how segment duration obeys prosodic conditioning comes from a comparison between the duration of the phonologically short, voiceless bilabial plosive in hobune “horse” and that of the corresponding long consonant in õpetaja “teacher”. Although the length of the first consonant is specified as Q1 and that of the second as Q2, the duration of these two consonants is identical according to Lehiste (1966) (70 ms).11 The reason for this lack of durational distinction 8

Lehiste, Consonant quantity and phonological units in Estonian, 16. Ibidem, 7. 10 For a comparative analysis of coarticulation in VC(C)V sequences in Japanese and Italian, see Caroline L. Smith, “The temporal organization of vowels and consonants”. Unpublished PhD diss., Yale University, 1992. For a comparison between Finnish and Italian), cf. Malcom H. Dunn, “The Phonetics and Phonology of Geminate Consonants: A Production Study”, Unpublished PhD diss., Yale University, 1993. 11 Lehiste, Consonant quantity and phonological units in Estonian, 6. 9

120

Chapter Six

between the two productions of /p/ probably lies in the fact that in õpetaja there is one more post-stress syllable, and also in the fact that the second syllable of that word is closed (in fact, the alveolar plosive is geminate); both these factors could easily explain why /p/ in õpetaja is not longer than in hobune.12 Additional suprasegmental factors may have an impact on segments’ surface durations, such as sentence stress (segments tend to be longer if included in prominent syllables within the intonational phrase) and the position in the phrase (both vowels and consonants lengthen in words at the end of the sentence).13 While Q1 and Q2 consonants are less affected by durational oscillations, Q3 consonants usually present larger variations, with durational values often overlapping those of the corresponding long consonants (or even of the short ones). Moreover, from a distributional point of view, the distinction between Q2 and Q3 for consonants is effective only when the consonant follows a short vowel; on the contrary, when the first syllable contains a vowel or a diphthong in Q3, the duration of the extra-long consonants does not substantially differ from that of the corresponding long consonants. On that basis one can argue that the only contrast admissible in that context is the one between a short and a generically long consonant. In addition, the possibilities of realizing consonant oppositions in contexts that differ from word initial disyllables are somewhat reduced. In monosyllables, the quantity of vowels and the quantity of consonants are in complementary distribution; only the alternation between Q1 and Q3 is realized. The types of sequence that are allowed in monosyllables are VVVCCC (Q3 + Q3, but with the above mentioned restriction on the generically long nature of the consonant in coda of a syllable after an extra-long vowel), VCCC or VVVC. The tendency to neutralize the opposition between Q2 and Q3, which has been noted for consonants following an extra-long vowel, is therefore valid, at least in monosyllables, also for vowels preceding an extra-long consonant.14 12 For similar results on Italian, where it is shown that the number of syllables in a word influences the duration of the segments in the stressed syllable, see Giovanna Marotta, Modelli e misure ritmiche. La durata vocalica in italiano (Bologna: Zanichelli, 1985). 13 Lehiste, Consonant quantity and phonological units in Estonian, 38. 14 Additional distributional constraints concern the word final position, in which the consonants may only be short (Q1) or long (Q2), as well as the word initial position, where consonant quantity is not contrastive. Furthermore, plosives in clusters may contrast at least two different degrees of length: cf. kirpu (with an /r/ + CC(C) sequence) vs. karbis (with /r/ + C) (the two labials measure 190 and 60

Geminate Timing in the Speech of Estonian L2 Learners of Italian

121

All of these factors clearly indicate that syllabic quantity governs the distribution of segments in Estonian, and that suprasegmental constraints overcome segment distributions in the three length degrees. The interaction of vocalic and consonantal quantity determines syllabic quantity (or “weight”), which supports the view of Estonian as a tendentially isomoraic language.15 The presence of a vowel or a consonant in Q3 is a necessary and sufficient condition to establish the extra-long quantity of the syllable. The presence of a further long (or extra-long) segment is redundant with respect to syllabic weight, though remaining (partially) contrastive on a segmental level. Moreover, it has been observed that the domain of quantity contrasts in Estonian mostly is the disyllabic foot. In fact, the durational variations of the disyllabic foot are relatively small compared to the variations of individual syllables (disyllabic foot isochrony). This produces a temporal balance within the disyllabic foot by which the duration of the first syllable increases from Q1 to Q2 to Q3 and, at the same time, the duration of the second syllable decreases. The tendency, though being styledependent to a certain extent (it has been reported to be more evident in read speech than in spontaneous conversation16), is a clear indication of how segmental quantity interacts with timing requirements within a trochee. In conclusion, the tendency to shorten vocalic segments before long consonants (i.e. in closed syllables) does exist in Estonian, as it does in Italian and in a large number of other languages. However, the phonology of Estonian allows long vowels to be realized before long or extra-long consonants, with the contextual restrictions seen above (e.g. in the case of monosyllables). In Italian, compensation phenomena are spared by any phonologization (as Italian phonology does not include quantity contrasts on vowels); on the contrary, they have a clearly allophonic nature. Moreover, compensation phenomena are highly variable, as has been shown by several studies. Italian is a tendentially isosyllabic language (though the picture has been found to be much more varied and complex than this traditional label ms, respectively, according to Lehiste, Consonant quantity and phonological units in Estonian,10). 15 Hayes, “Compensatory lengthening in moraic phonology”, 253-306; Paul Kiparsky, “Livonian stød”, in Segments and Tone eds. Paul Boersma et al. (Tübingen: Niemeyer, 2006), 1-16. 16 Diana Krull, “Foot isochrony in Estonian”, in Proceedings of the XIV International Congress of Phonetic Sciences (ICPhS 99) eds. John J. Ohala et al. (San Francisco: University of California, 1999), 1063-66.

122

Chapter Six

suggests). A stressed vowel preceding a geminated consonant or a consonant cluster tend to be shorter compared to corresponding vowels preceding a singleton consonant, all other conditions being equal. However, compensation is particularly evident when vowels are in a prominent position in the intonational phrase and/or in isolated words,17 the process is also influenced by lexical-prosodic factors such as the number of following unstressed syllables.18 The analyses of C-to-V ratio carried out by Bertinetto and Vivalda (1978) and by Pickett et al. (1999) have revealed that compensatory effects are effectively salient for the speaker-hearer only in those contexts in which the consonant does not exceed a given duration threshold; in the contexts in which the geminate consonant does exceed the duration threshold, vowel compensation becomes perceptually no longer informative, and the distinction between singleton and geminate consonant is recovered by speakers on the sole basis of consonant duration.

2.2. Hypotheses and objectives This research project aims to examine the procedures of integration and adaptation of Italian gemination patterns in the interlanguage of Estonian learners. In particular, based on cross-linguistic comparisons in the existing literature on geminate timing, we focus on the following two aspects: the geminate-to-singleton consonant duration ratio (see below, § 4.1.) and the duration ratio between the preceding vowel and the entire VC sequence (see below, § 4.2.). Concerning the first measure, the above-mentioned study by Ham (2001), which was carried out on a sample of 14 languages, revealed that geminate consonants were longer in languages based on moraic isochrony than in languages based on syllabic isochrony (compared to the corresponding singleton consonants19). According to this generalization, we should find that the geminate-to-singleton consonant duration ratio is greater in Estonian speech than in Italian speech (consonant type and 17

Giovanna Marotta, Modelli e misure ritmiche. La durata vocalica in italiano (Bologna: Zanichelli, 1985). 18 Pier Marco Bertinetto, Strutture prosodiche dell’Italiano (Firenze: Accademia della Crusca, 1981); Marotta, Modelli e misure ritmiche. La durata vocalica in italiano; Kristie M. McCrary, “Reassessing the Role of the Syllable in Italian Phonology: An Experimental Study of Consonant Cluster Syllabification, Definite Article Allomorphy and Segment duration”, Ph.D. diss., UCLA, 2003. 19 William H Ham, Phonetics and phonological aspects of geminate timing (London: Routledge, 2001), 163.

Geminate Timing in the Speech of Estonian L2 Learners of Italian

123

phonotactic context being equal). Moreover, from the acquisitional point of view, we will verify whether and how geminate timing changes in the interlanguage as phonological acquisition advances. The first experiment therefore analyses the geminate-to-singleton consonant duration ratio in the interlanguage of Estonian learners of Italian, compared to the duration ratios of Italian and Estonian as L1s. With regard to the second measure, i.e. the duration ratio between the preceding vowel and the entire VC sequence, the study by Ham (2001) shows (based on a sample of 4 languages20) that the information content of timing relationships in VC sequences tends to be higher for isosyllabic languages than for isomoraic languages. This may be due to the fact that isomoraic languages like Estonian impose fewer restrictions of phoneme concatenation compared to languages like Italian, allowing sequences of two long (or extra-long) segments. Indeed, the disyllabic foot isochrony which is found in Estonian could be a specific manifestation of the tendency to underemphasize timing relations between elements composing the VC sequence, since it introduces a balancing effect precisely within the foot. Estonian and Italian should therefore show a different behaviour with respect to the durational relation between a stressed vowel and the following consonant, Italian being more sensitive than Estonian to the temporal variations of the consonant. From an acquisitional point of view, it will therefore be possible to verify if, and to what extent, a greater sensitivity to the timing relationships within the VC sequence develops in the interlanguage as phonological acquisition advances. The second experiment analyses the V-to-VC duration ratio in the interlanguage of Estonian learners of Italian, compared to Italian and Estonian as L1s.

3. Experiments 3.1. Materials and methods The speech material is made up of a series of recordings of the Corpus of L2 Italian of the University for Foreigners Perugia,21 and in particular

20 These are a rural Bernese dialect (Switzerland), Levantine Arabic, Madurese (an Austronesian language from the Indonesian group) and Hungarian. 21 The corpus of Italian L2 has been collected since the 1990s at the University for Foreigners Perugia and contains interviews, spontaneous and semi-spontaneous conversations, readings, repetitions and a series of perceptual tests produced by about 300 students of various mother tongues and different levels of linguistic competence in L2 Italian. For the purposes of the present research, a subsection of

124

Chapter Six

of the speech of four young native speakers of Estonian, enrolled in the University for Foreigners language and civilization programmes. At the time of the recording, they all had low proficiency in Italian, roughly corresponding to A1 or A2 learners according to the Common European Framework of Reference for Languages (CEFR). Both read and repeated speech were analysed. In the first case, the students had to read isolated words and short sentences in Italian, presented on a paper sheet in a university classroom and in the exclusive presence of the interviewer. In the second case, the students listened to isolated words produced by a native speaker of Italian and they were asked to repeat them aloud.22 Only the productions where no error occurred were retained for analysis; specifically, correct productions were considered whenever the learner (i) showed that he or she had understood the gross meaning of the sentence or word produced; (ii) did not replace, insert or omit any segment or syllable; (iii) pronounced lexical stress correctly; (iv) did not make pauses, hesitations or expressive extensions. For the purpose of the comparison with L1 Italian, a young female native Italian speaker of Umbrian origin was asked to read the same list of words and short sentences.

the corpus is currently undergoing digitization and archiving at the Linguistics Laboratory of the Scuola Normale Superiore of Pisa. 22 We are well aware of the fact that any production elicited through a repetition task is the expression of a process of decoding + imitating the auditory signal that, by its own nature, cannot be considered a faithful and immediate reflection of the speakers’ lexical intention. The auditory model supplied by the stimulus influences directly the production of the stimulus itself (although we have to specify that there were no temporal constraints for the subjects to produce the stimulus, and the learners often repeated the word only when they were sure they had understood its meaning). For this reason, we preferred to use read speech rather than repeated speech, resorting to the latter only in cases of scarcity of data relative to a subclass of words or a segmental category, and only with reference to terms and expressions that were highly familiar to the learner and whose repetition was produced without errors or uncertainties. On the other hand, it must be said that read speech is also affected by “external” influences on verbal production, such as the influence of orthography (and consider that, for L2 learners, there is the additional problem of the orthographic interference of L1 over L2 norms). Spontaneous speech could overcome similar problems. Unfortunately, the subjects of this study were beginner learners of Italian and spontaneous speech data – though present in the Perugia corpus – did not prove to be quantitatively and qualitatively adequate. We reserve for future research the analysis of spontaneous or semi-spontaneous speech to verify whether the generalizations proposed in this study for read (and repeated) speech could be extended to more natural speech styles and conditions.

Geminate Timing in the Speech of Estonian L2 Learners of Italian

125

3.2. Analysis In this analysis we assume that the temporal domain is the most informative domain for consonant gemination (compared to other domains, such as the frequency spectrum). This assumption is motivated by a large body of acoustic studies, according to which duration is the primary factor differentiating singleton and geminate consonants in Italian23 and in several other languages.24 In particular, as far as plosives are concerned, it has been shown that the most informative parameter is occlusion duration, while release and VOT durations do not carry significant distinctions.25 Typically, in the study of geminates, one or more of the following ratios are analysed: CC/C (geminate-to-singleton ratio), C/V (ratio between a singleton or geminate consonant and the preceding vowel), V/V+C (ratio between the preceding vowel and the VC/VCC sequence). Language-specific timing patterns allow the listener to perceptually distinguish the singleton vs. geminate consonant contrast

23 Pier Marco Bertinetto and Enrico Vivalda, “Recherches sur la perception des oppositions de quantité en italien”, Journal of Italian Linguistics 3 (1978): 97-116; Loredana Cerrato and Mauro Falcone, “Acoustic and perceptual characteristic of Italian stop consonants” (paper presented at the 5th International Conference on Spoken Language Processing (ICSLP), Sydney, Australia, November 30 December 4, 1998, paper 0463); Anna Esposito and Maria Gabriella Di Benedetto, “Acoustic and perceptual study of gemination in Italian stops”, Journal of the Acoustical Society of America 30 (1999): 175-85; Emily R. Pickett, et al., “Effects of Speaking Rate on the Singleton-Geminate Consonant Contrast in Italian”, Phonetica 56 (1999): 135-57. 24 For Swedish, see Peter E. Czigler, “Timing in Swedish VC(C) Sequences”, Reports from the Department of Phonetics Umeå University, PHONUM 5, Umeå, Sweden: Department of Phonetics, Umeå University, 1998. For Persian see Benjamin B. Hansen, “Production of Persian Geminate Stops: Effects, of varying speaking rate”, in Proceedings of the 2003 Texas Linguistics Society Conference, eds. Augustine Agwele, Willis Warren and Sang-Hoon (ParkSomerville, MA: Cascadilla Proceedings Project, 2004), 86-95. For Tashlhiyt Berber, see Rachid Ridouane, “Geminates at the junction of phonetics and phonology”, in Papers in Laboratory Phonology 10, eds. by Cécile Fougeron et al., 61-80. (Berlin / New York: De Gruyter, 2010). For various languages, see Ham, Phonetics and phonological aspects of geminate timing. 25 Esposito and Di Benedetto, “Acoustic and perceptual study of gemination in Italian stops,” 175-85. On the other hand, VOT duration notoriously is a clue for place and voicing distinctions among plosives.

126

Chapter Six

across individual variations (i.e. for different voices) and stylistic conditions (i.e. at different speech rates).26 The segmentation, annotation (both phonetic/phonemic and syllabic) and acoustic analysis are still in progress. The software used is Praat 5.1.32. The duration of all the intervocalic consonants (both singleton and geminates, in stressed and unstressed syllables) and of the preceding vowels is calculated with a Praat script purposely created.27 For the aims of this study, the corpus has been balanced for the number of singleton and geminate consonants across segment types (plosives, fricatives, affricates, nasals and liquids), places of articulation, voicing degrees and stress condition (pre-tonic vs. post-tonic).28 The final corpus comprised 646 consonants and 406 vowels produced by the Estonian learners, plus 154 consonants and 146 vowels produced by a native speaker of Italian. The results reported on in the following sections deal with plosives (both voiced and unvoiced), liquids and nasals. In the first experiment (see below, § 4.1.), the duration ratio between the singleton and the corresponding geminate consonant was calculated (for plosives, separately for the occlusion phase and the total duration of the phone). In the second experiment (see below, § 4.2.), only disyllabic paroxytones were retained; the duration ratio between a stressed vowel and the whole sequence comprising the stressed vowel and the following consonant was calculated.

4. Results 4.1. First experiment: geminate-to-singleton ratio The data in Lehiste (1966), though being based on the production of only one speaker, still represent an important contribution to the study of consonant duration in Estonian. In particular, it can be utilized as a point of comparison with the acquisitional data in our possession because the elicitation procedure was similar. The stimuli were short sentences or 26

Emily R. Pickett, Sheila E. Blumstein and Martha W. Burton, “Effects of Speaking Rate on the Singleton-Geminate Consonant Contrast in Italian,” Phonetica 56 (1999): 135-57. 27 We gratefully acknowledge Chiara Bertini (Scuola Normale Superiore, Pisa) for having provided the script. 28 Preceding vowels were not balanced for quality: a variable number of high, low, mid-high and mid-low vowels appeared in the experimental list. However, given that the experimental list was the same for the learners and the native Italian speaker, vowel quality was balanced across subject groups.

Geminate Timing in the Speech of Estonian L2 Learners of Italian

127

isolated words; they were produced through reading aloud. In the execution of the project, first hand data will be collected specifically to be integrated into the Lehiste’s corpus and provide more evidence on consonant quantity patterns in use in L1 Estonian. Table 6-1 contains the average duration of intervocalic consonants between the first and second syllable of polysyllabic words (post-tonic position), according to Lehiste (1966). The author did not specify whether the occlusion phase or the entire phone was measured; we hypothesize that the entire phone was consistently measured. There is also no distinction based on the duration of the preceding vowel; all consonants are considered cumulatively and independently from the prosodic structure of the foot. First, it can be noted that the duration ratio between singleton (Q1) and geminate consonants (Q2) is rather low, which indicates that geminates are much longer than the corresponding singleton consonants (in most cases, they measure more than twice the corresponding singleton). Second, different places of articulation may produce different timing conditions. Consider in particular coronal plosives, which are the target of the largest variation in the three quantity distinction, with an average duration of 30, 83 and 152 ms for Q1, Q2 and Q3, respectively. Concerning sibilants, we may see that the singleton-to-geminate ratio is higher than in plosives, thus indicating that the durational differences between Q1 and Q2 sibilants are smaller.

C/CC C/CCC CC/CCC

/p/

/t/

/k/

1:2 1:2.9 1:1.5

1:2.7 1:5 1:1.8

1:2.1 1:4 1:2

plosives (average) 1:2.3 1:4.9 1:2.2

/s/ 1:1.4 1:3.1 1:2.1

Table 6-1: Duration ratios between intervocalic consonants in Q1 (C), Q2 (CC) and Q3 (CCC) in L1 Estonian (average values calculated for the data in Lehiste 1966). Table 6-2 contains the data of L1 Italian as produced by our reference speaker. The table is divided into two parts, one containing the duration of the occlusion phase, the other containing the duration of the phone. Only intervocalic consonants preceded by tonic vowels were taken into consideration. Voiced velar plosives were not included in the analysis since only a few cases were present in the dataset.

Chapter Six

128

/p/ Duration 1:1.2 of the occlusion phase Duration 1:1.1 of the entire phone

/t/

/k/

/b/

/d/

plosives /l/ average 1:1.8

1:1.9 1:1.5 1:2.6

1:2

1:2.1 1:1.4 1:2.6

1:1.8 1:1.8

1:3.3

/r/

/m/

/n/

1:3.3 1:2.4 1:3.2

Table 6-2: Singleton-to-geminate duration ratios in L1 Italian. According to Table 6-2, geminate plosives in Italian are less than twice as long as the corresponding singletons. Moreover, variations were present according to place of articulation (bilabial, alveodental, velar) and voicing. In the case of the voiceless bilabial, the duration of the geminate is only 1.1 or 1.2 times greater than the duration of the singleton consonant, while in the case of the voiceless alveolar, the geminate was approximately twice as long as the singleton counterpart. The behaviour of velars lies somewhere between the two. On the other hand, for voiced consonants the duration ratio is significantly smaller than for voiceless consonants, thus indicating that the durational difference is greater for the voiced than for the voiceless series. Considering the duration of the occlusion phase or of the entire phone does not introduce significant changes in the overall picture. Finally, continuous consonants show a substantially different pattern compared to plosives, with considerably lower duration ratio values. Several studies have addressed the topic of quantity relations in L1 Italian. Among them, the study by Bertinetto and Vivalda (1978) showed that, in the case of the voiceless bilabial plosive and the voiceless labiodental fricative, the singleton-to-geminate duration ratio is on average 1:1.87. The study by Esposito and Di Benedetto (1999) reported on a very similar value, i.e. 1:1.90, for the entire series of plosives. In the Pickett et al. (1999) study, the slightly higher value of 1:2 for voiceless bilabial and alveodental plosives was reported. Finally, according to Ham (2001), the two studies by Smith (1992) and Dunn (1993) reported on a ratio value of 1:1.85. In all these studies, the duration of the occlusion phase was measured. The study by Cerrato and Falcone (1998) took into consideration the duration of the whole segment (thus including the burst and the

Geminate Timing in the Speech of Estonian L2 Learners of Italian

129

release) and the resulting duration ratio was equal to 1:1.81.29 In conclusion, as far as plosives are considered, the data obtained from the reference native Italian speaker and those attested in the literature appear to agree substantially. Table 6-3 contains the results of the analysis of the speech of the Estonian learners. Again, in this case, the duration ratios were calculated separately for the occlusion phase and for the entire phone; however, the results show that there is no substantial difference in adopting one or the other analysis window. Place of articulation and voicing turned out to introduce wide variation in ratio values. Similar to what was found for the L1 Italian native speaker, voiced geminates tended to be longer (compared to the corresponding singletons) than voiceless geminates. On the other hand, the speech of the Estonian learners also presented partially different timing patterns with respect to the Italian reference speaker. In particular, it was the velar rather than the coronal plosive that showed the lowest ratio value.

Duration of the occlusion phase Duration of the entire phone

/p/

/t/

/k/

/b/

/d/

plosives (average)

1:1.3

1:1.2

1:1.9

1:2.4

1:2

1:1.8

1:1.1

1:1.3

1:1.8

1:2.3

1:2.1

1:1.7

/l/

/r/

/m/

1:2.8

1:5 1:2.5

/n/

1:3.4

Table 6-3: Singleton-to-geminate duration ratios in the L2 Italian of Estonian learners. According to our informal observations on the speech of the Estonian learners, the coronal plosive tended to be pronounced as palatal. If this tendency is to be supported by more specific data, it could contribute to the explanation of the small difference existing between the duration of the singleton and of the geminate consonants in the speech of the learners. Future research will be devoted to the analysis of the articulatory 29

The ratios have been calculated by the authors of this paper on the basis of the average duration values reported on in the various experiments. The authors therefore take the responsibility for the accuracy of the calculations and the reliability of the ratio values.

130

Chapter Six

characteristics of individual consonants in the Estonian-Italian interlanguage. Additional peculiarities were noticeable in the production of the geminate velar plosives, which sounded particularly long in the Estonian pronunciation. Continuous consonants in the speech of the learners showed values for geminate lengthening greater than those of the plosives, consistent with native pronunciation patterns. Rhotics, however, varied and presented extremely lengthened geminates, compared to the other sound classes. It should be noted that singleton rhotics were uniformly produced as taps by the learners (not so by the native Italian speaker, who often produced short trills). On the other hand, geminate rhotics were most often produced as trills with a very high number of contacts, even up to 5 or 6. An example of long trill produced with 5 linguo-palatal contacts is reported in Fig. 6-1.

Fig. 6-1: Oscillogram and spectrogram of [‫ޖ‬mir:a] mirra (“myrrh”) in the speech of an Estonian learner.

Geminate Timing in the Speech of Estonian L2 Learners of Italian

131

4.2. Second experiment: V-to-VC duration ratio The V-to-VC duration ratio is usually investigated30 to obtain information on the percentage of the temporal interval occupied by the vowel within the VC sequence. Therefore, it provides indirect clues on the presence of compensatory effects on the vowel preceding a long consonant. It has been noted that, like the singleton-to-geminate duration ratio, the V-to-VC duration ratio remains stable across speakers and contexts, in English31 and various other languages.32 It therefore serves the purpose of differentiating singletons from geminate consonants in perception. At this stage of the research, since data for the analysis of VC timing patterns in L1 Estonian are still unavailable, we will focus on L1 and L2 Italian only. Among the priorities of this research is the analysis of VC timing patterns for a sufficiently ample corpus of native Estonian speech. Table 6-4 contains the durational values of the vowel-consonant sequences and the corresponding values of the V-to-VC ratio in the speech of the reference native Italian speaker, while Table 6-5 presents the results for the Estonian learners, separately for voiceless and voiced plosives, liquids and nasals. Only paroxytone disyllables were included. The average values of Table 6-4 clearly indicate that, in native Italian speech, a stressed vowel before a geminate consonant did not undergo any shortening compared to the corresponding vowel before a singleton consonant. In some cases, consonant lengthening appeared to be balanced by a comparable lengthening of the vowel. The two most striking cases are those of the /Vtt/ and /Vkk/ sequences, but it may be said that in general this tendency applies to the entire plosive group. Consequently, the V-toVC duration ratio values were not substantially different across contexts. Liquids and nasals showed a different picture, as gemination provoked a substantial lengthening of the consonant which was not reflected in preceding vowel. The latter tended to have the same average duration in the two contexts, and for this reason occupied a more reduced portion of the VC sequence when followed by a geminate. Though the data were taken from a small number of productions of a single speaker, the tendency proved consistent enough with what we know about the variability of the compensatory mechanism in closed syllables in Italian (cf. above, § 2.1.). 30 Klaus J. Kohler, “The production of plosives”, Arbeitsberichte des Instituts für Phonetik der Universität Kiel (AIPUK) 8 (1977): 30–110. 31 Robert F. Port, “Linguistic timing factors in combination”, Journal of the Acoustical Society of America 69 (1981): 262-74. 32 Ham, Phonetics and phonological aspects of geminate timing.

Chapter Six

132

VC Vp Vt Vk Vb Vd Vl/Vr Vm/Vn

V duration C duration V-to(ms) (ms) VC ratio 210 137 0.61 174 118 0.60 173 119 0.59 167 105 0.61 159 95 0.62 199 50 0.8 188 98 0.65

VCC Vpp Vtt Vkk Vbb Vdd Vll/Vrr Vmm/Vnn

V duration CC (ms) duration (ms) 199 197 275 132 203 178 198 225 167 154 181 223 199 213

V-toVC ratio 0.5 0.67 0.53 0.47 0.53 0.45 0.48

Table 6-4: Duration of the VC sequences and corresponding V-to-VC duration ratio in L1 Italian. VC

duration V Vp 132 Vt 191 Vk 126 Vb 183 Vd 189 Vl/Vr 198 Vm/Vn 197

duration C 163 187 203 81 86 86 94

V/V+C VCC 0.46 0.51 0.39 0.7 0.68 0.7 0.68

Vpp Vtt Vkk Vbb Vdd Vll/Vrr Vmm/Vnn

duration V 80 90 99 106 124 127 134

duration CC 262 258 264 275 236 210 272

V/V+C 0.23 0.25 0.27 0.27 0.34 0.38 0.33

Table 6-5: Duration of the VC sequences and corresponding V-to-VC duration ratio in the L2 Italian of Estonian learners. The acquisitional data present a quite different picture (see Table 6-5). In both cases of plosives, liquids and nasals, a substantial shortening of stressed vowel was found in conjunction with consonant gemination. This pattern appeared to be followed by all consonant subgroups without exception. The V-to-VC duration ratios were consistently different in the context of singleton consonants compared to those with geminates for both liquids and nasals, as well as for both voiceless and voiced obstruents. The Estonian learners, therefore, appeared to produce the vowel balance effect consistently for all consonant classes, independently from the degree of lengthening that different types of geminate introduced on a segmental level. Furthermore, we have to recall that the singleton-to-geminate duration ratio in L2 Italian showed a significant difference between the group of plosives and the group of continuants, as only the latter produced substantial lengthening on geminates (see above, § 4.1. and Table 6-3).

Geminate Timing in the Speech of Estonian L2 Learners of Italian

133

Nevertheless, stressed vowels in disyllables were always indiscriminately shorter in percentage on the duration of the entire VC sequence, when followed by a geminate compared to when followed by a singleton consonant.

5. Discussion The two experiments described here have analysed the temporal aspects of segments (first experiment) and of the VC sequence (second experiment) with respect to the phenomenon of consonant gemination in the L2 Italian of Estonian learners. L1 Italian and—whenever possible— L1 Estonian speech were also analyzed for purposes of comparison. In the first experiment, the Estonian learners seemed to produce a pattern of singleton-geminate alternation close to that of the native speakers of Italian, as the available literature as well as the measurements carried out on the reference speaker have shown. The average value of the singletonto-geminate duration ratio for plosives produced by the learners is around 1:1.8 (considering only the plosive phase), while it oscillated between 1:1.8 and 1:2 for the native Italian speakers. In both populations, geminate lengthening was much greater for the class of continuous consonants (liquids and nasals) than for plosives. On the other hand, the value of the singleton-to-geminate duration ratio in native Estonian speech proved to be substantially inferior, with average values for plosives around 1:2.3 (see above, § 4.1.). This seems to suggest, in agreement with Ham’s (2001) prediction of a difference existing between isosyllabic and isomoraic languages, that Estonian has comparatively longer geminates than Italian. With respect to this parameter, the Estonian learners’ interlanguage seemed to be placed in a position more similar to that of the target language than to that of Estonian. The second experiment showed more substantial differences concerning stressed vowel durations between native Italian and non-native speech. However, given the lack of data for L1 Estonian, the analysis was based exclusively on the comparison between the Estonian learners’ interlanguage and the L1 Italian of a single reference speaker. We saw that, in L1 Italian, stressed vowels before geminates occupied a smaller interval of the VC sequence (compared to before singleton consonants) when the consonant is a continuous consonant, while there was no difference between sequences with geminate and sequences with singleton plosives (see above, § 4.2.). We must consider that the analysed speech is made up of a list of words and short sentences, thus a form of production that should favour the application of the compensatory

134

Chapter Six

mechanism;33 nevertheless, plosive gemination did not seem to provoke any significant change in the temporal pattern of the stressed vowel with respect to the VC sequence. Ham’s (2001) hypothesis on the high information content of the VC sequence in isosyllabic languages does not seem to be fully supported by the Italian data. However, further testing is needed to substantiate these hypotheses. In the second experiment, the Estonian learners produce vowel compensation in a consistent way on all consonant classes, independently of the degree of the lengthening that the various types of geminate introduce on a segmental level. It should be remembered that the singleton-to-geminate duration ratio in the L2 Italian of our Estonian learners showed a difference between plosives (where geminates are only slightly longer than the corresponding singletons) and continuous consonants (where lengthening was definitely more consistent; see below, § 4.1. and Table 6-3). Nevertheless, stressed vowels in disyllables were always indiscriminately shorter in percentage on the duration of the entire VC sequence, when followed by a geminate compared to when they are followed by a single consonant. There are various cross-linguistic implications for this type of result. The VC timing patterns in Estonian learners’ speech seem to emphasize the compensatory variations of the vowel compared to native Italian speech. It cannot be said, at this stage of the research, whether there is an influence of the learners’ native language, or rather an emphasis given to a feature perceived as a means of expression of the consonant duration distinction in the target language. In the latter hypothesis, and considering that Estonian is a tendentially isomoraic language, the generalisation of vowel compensation might reflect the transition from a system based on the quantity of the disyllabic foot, in which sequential timing plays a marginal role in the expression of phonological contrasts, to a system more sensitive to syllabic timing. According to this hypothesis, the Estonian learners would be introducing segmental balancing relationships within the moraic trochee.

6. Conclusions The research project presented here opens several possibilities for further investigation. Although the two experiments require additional verification and a substantial increase of data analysis, they apparently 33

See in particular Marotta, Modelli e misure ritmiche. La durata vocalica in italiano.

Geminate Timing in the Speech of Estonian L2 Learners of Italian

135

show that the process of restructuring of geminate timing in the interlanguage of Estonian learners is a composite phenomenon involving both levels of segmental and of VC quantity. In learners’ restructuring of native prosodic patterns, phonological factors and phonetic tendencies differently characterizing the Estonian and Italian systems appear to interact in a complex way. For example, our Estonian learners adjusted the duration contrast on a segmental level by introducing a singleton-to-geminate ratio closer to that of native Italian speakers, at least as far as plosives are concerned. On the one hand, the effects of the native language articulatory characteristics appeared to rule over the production of continuous consonants, especially rhotics. On the other hand, the interlanguage varied to a large extent with respect to the application of a compensatory vowel shortening within the trochaic disyllable. It is also possible that learners find strongly variable cues of vowel compensation in the input of the linguistic environment, thus reflecting the input variability in the phonological restructuring of the interlanguage. All these points are worth being addressed with a larger numbers of speakers (both Estonian learners and native Italian speakers) and, where possible, with respect to different speaking styles. Furthermore, future research will also require a detailed analysis of consonant and vowel quantity in L1 Estonian, in order to allow cross-linguistic comparisons.

CHAPTER SEVEN A RHYTHMIC-PROSODIC ANALYSIS OF L1 AND L2 ITALIAN LUCIANO ROMITO UNIVERSITY OF CALABRIA, ITALY

AND ANDREA TARASI UNIVERSITY OF SALERNO, ITALY

1. Introduction In Italy, acquisitional linguistic studies are a recent development and generally concentrate on the morpho-syntactic and textual level of an L2. In recent years, studies have shown that the prosodic level is subject to a transfer from the first language to the second, and that the phonological level is that in which the influence of L1 is more obvious and more difficult to conceal. L2 learners can easily acquire a good knowledge of different levels of a language but will always find it very difficult to control rhythm and prosody. They can achieve competence in terms of the phonological system of the L2, pronouncing single words in a form almost indistinguishable from that of native speakers, but when they produce longer sentences or phrases, their identity as foreigners emerges immediately. What the L1 depends on is the entire system, which determines the rhythm, intonation and accentuation of the sentence—in short, the process of linguistic transfer. The spread of new methods for assessing the rhythmic aspects of languages has provided a strong new impetus to research in this field. It has been noted that the assessment of the effects associated with a different rhythmic-temporal organization of speech production in various languages depends on the segmentation of acoustic data and on their evaluation in phonetic and phonological terms. In addition, numerous studies have shown that the difference between isosyllabic and

138

Chapter Seven

isoaccentual languages does not produce strong experimental feedback. Only recently has the measurement of some variables been related to phonological and syllabic temporal properties and the accentual structure of the various languages. Also, conditions associated with specific phonetic realizations have appeared to allow the detection of acoustic parameters capable of identifying the rhythmic properties of languages. Italian, unlike other European languages with a more consolidated and ancient history, has developed a broad diversification in which specific varieties can be recognized. Variation is related to the dimension of all extra-linguistic parameters, such as diatopic variation (the geographic area where the language is used), diastratic variation (stratum or social group to which speakers belong), and diaphasic variation (communicative situation in which language is used). These dimensions are the reference axis along which varieties can be simultaneously ordered in the space of variation in contemporary Italian and also in those produced by foreign speakers. This is extremely important in evaluating an L2 Italian. It should not be forgotten that, in common use, these dimensions intersect, and the varieties can move simultaneously over axes of variation.

2. Materials and Methods In this study, we consider spontaneous acquisition only. This contact with the second language takes place exclusively in situations of natural verbal exchange for communicative purposes and to satisfy basic needs. The learner may have greater control at the pragmatic level of a language, while his/her language skills may be less extensive, and in some cases limited to an elementary level. We used a multilingual corpus (Pangea) containing linguistic codes for L1 and L2 Italian composed of 25 speakers of different sex, school ranking and age. Furthermore, the speakers had been living in Calabria for a minimum of approximately two years. The languages considered in this study were: Albanian (3 M and 2 F aged 2030), Chinese (3 M and 1 F aged 20-30) Italian (4 M aged 35-70), Polish (2 M and 3 F aged 20-55), and Romanian (5 F aged 20-50). The L1 languages are typologically different, and the subjects read a list of sentences and also produced at least three minutes of spontaneous speech in which the topic was chosen by the speakers themselves so as to guarantee greater spontaneity. We considered both spontaneous and read speech in the analysis. Recordings were made with an Edirol 24-bit digital

A Rhythmic-Prosodic Analysis of Italian L1 and L2

139

recorder and a Sennheiser microphone. The signal was labelled1 using Praat software.2 The parameters analyzed for spontaneous speech were articulation rate, considered specifically for each language,3 the duration of unstressed and stressed syllables, and related components. The values obtained from these parameters were normalized and inserted into a figure which represents the Italian Index. Studies on rhythm were conducted using PVI,4 CCI,5 and GVI.6 On read speech we proposed a perception test in which the participants had to discriminate L1 Italian from L2 Italian, a musical analysis where the sentences are transferred to a pentagram and, finally, we tested an automatic speech recognition system called Trascrivi.it, developed by Cedat85. This is a trade system based on general models for standard Italian without a specific form of training. In this case, we analyze the system’s capability to recognize and distinguish L1 from L2 Italian. We discuss the results obtained from: perception and musical analysis, automatic SR, acoustic analysis and rhythmical index. Following these analyses, we propose an index called Italian Index which plays a dual role. On the one hand, it pursues the scientific objective of verifying the rhythmic characteristics that exist in a language, in this case the differences between L1 Italian produced by Calabrian speakers and L2 Italian produced by Polish, Romanian, Chinese and Albanian speakers. On 1

Luciano Romito and John Trumper, “Problemi teorici e sperimentali posti dall’isocronia”, in Quaderni del Dipartimento di Linguistica dell’Università della Calabria, ed. Cattedra di Linguistica, (Cosenza: University of Calabria, 1993), 88115. 2 Syllables that present phonetic deletions, strong reduction or centralization were not considered for the calculation of the indices, but the treatment is different for Articulation Rate. Syllable lengthening was not considered in calculating the parameters and their duration was eliminated from total duration of the voice recording. 3 Massimo Pettorino, “La velocità di articolazione”, in Costituzione, gestione e restauro di corpora vocali, eds. Amedeo De Dominicis et al. (Roma: Esagrafica, 2003), 227-32. 4 Esther Grabe and Ee Ling Low, “Durational variability in speech and the rhythm class hypothesis”, in Papers in Laboratory Phonology 7, ed. N. Warner (Berlin: Mouton de Gruyter, 2002), 515-46. 5 Pier Marco Bertinetto and Chiara Bertini, “On modeling the rhythm of natural languages” (paper presented at the biennal meeting for Speech Prosody, Campinas, São Paulo, May 6-9, 2008). 6 Joseph Tepperman and Emily Nava, “long-distance rhythmic dependencies and their application to automatic language identification” (paper presented at the annual conference for Interspeech, Florence, Toscana, August 28-31, 2011).

140

Chapter Seven

the other hand, it pursues an educational purpose since it identifies the critical points in speech produced by L2 Italian speakers, thus enabling the identification of potential educational approaches aimed at filling the gaps between LI and L2 Italian.

3. Data Analysis In this paragraph, we present the data obtained from the analysis carried out on spontaneous and read speech. The languages have a different syllabic structure, and in many experimental works it has been shown that a greater or lesser syllabic complexity can have a significant impact on the progress of the rhythm in languages. The average length of a syllable due to the different percentage of open and closed syllables and the number of segments influence some parameters such as, for example, articulation rate. The values of this macro-prosodic index thus tend to be low in languages with a complex syllable structure and high in languages with a relatively simple one, such as Italian. Results of the perception test carried out on read speech demonstrate the significant influence, as far as regards language discrimination, of the phonetical and phonological systems of the first language, as can be seen in Fig. 7-1. Table 7-1 exemplifies the importance, in percentage terms, of the linguistic levels for each language.

Fig. 7-1: Results of the Perception Test.

A Rhythmic-Prosodic Analysis of Italian L1 and L2

Seg. Lev. Supraseg. Lev.

141

Alb

Chin

Pol

Rom

South It

Nort h It

79.31% 1.72%

17.24% 32.76%

52.63% 19.30%

36.84% 17.54%

87.10% 3.23%

85.19% 14.81%

Table 7-1: Results in percentage of the perception test. The only relevant data for the prosodic level are the 32.76% for L2 Italian produced by the Chinese speakers. In this case, the segmental level is clearly superior in perceptive terms with respect to the role played by the prosodic level. In effect, in this model, intonative-metric analysis7 shows that L1 Chinese assigns different prominences, defining a different intonation pattern. The Chinese speaker places his communicative focus on different lexical elements compared to the L1 Italian speaker, because he considers other lexical elements to be important. The Chinese speaker gives prominence to the last syllable of the whole sentence “[…] base di ciascun raggruppamento di fonemi” which underlines the speaker’s desire to emphasize this part of the sentence producing, in agreement with this element, a strong prosodic foot. The differences between the prosodic and the syntactic trees produce a series of odd variations, and the Chinese speaker creates three intonative phrases, the last of which is the most marked and strong. So, in a manner different from the Italian model, the last part of the sentence conveys the information. There are also differences on the syntactic tree because the speaker creates two pauses, generating a different syntactic pattern. The Italian speaker produces a pause before the relative clause, while the Chinese speaker pauses before the verb and the relative clause, emphasizing that part of the sentence. In effect, it is the strong intonation phrase that conveys the information. Before the pauses, the Chinese speaker manifests rising intonation that represents an intonative continuum, which involves a temporal lengthening of the syllable. Probably, in this case, there is a transfer from his L1 to his L2 Italian. Figure 7-2 represents the pitch contour of L1 Chinese. From the phonetic and phonological point of view, listeners identified different phenomena: the loss of geminate consonants (for example the Italian term raggruppamento becomes [ragrupa‫ޖ‬mento]) because the phonological opposition C~CC does not exist in the considered L1. Some phones are in a manner different to L1 Italian (for example L1 Albanian 7

Luciano Romito, et al, “Analisi percettiva, musicale e automatica dell’italiano L1 e L2” (paper presented at the annual conference for the A.I.S.V., Rome, Italy, January 25-27, 2012, in press).

142

Chapter Seven

speakers make =l? alveolar-approximant while in Italian it is lateralalveolar). An unvoiced sound in some cases becomes voiced (for example L1 Chinese produces the Italian phone =p? as =b? because the first is absent in its phonological system). These cases clearly represent a linguistic transfer from their first language to their Italian as a second language.

Fig. 7-2: Pitch contour for L1 Chinese.

Obviously, the importance of the phonetic-phonological level is also proved by the results obtained from the SR system8 where, without a semantic control, the substitution of the phones occurs acoustically and induces the system to search for words out of context (for example, the results returned by the system for L2 Italian of Chinese speakers is “la sera per sono le strutture più elementari castano come passati ciascun accorpamento che fonte fonemi”, while the produced sentence is “le sillabe sono le strutture più elementari che stanno come base di ciascun raggruppamento di fonemi”.). The system uses word recognition on a prosodic basis to distinguish the words and finally uses the phonetic system for their recognition. A mistaken interpretation causes the system to recognize the sentence, as can be seen in the preceding example. The system will always produce an Italian word, without, however, performing a morphological, syntactic and semantic analysis of the sentences. All L2 8

Romito et al., “Analisi percettiva, musicale e automatica dell’italiano L1 e L2”.

A Rhythmic-Prosodic Analysis of Italian L1 and L2

143

Italian speakers give greater value to stressed syllables (SS) compared to unstressed syllables (US) (Table 7-2). Alb SS US

0.20 0.14

Chin

Pol

0.24 0.18

Rom

0.22 0.16

0.20 0.14

Ita 0.22 0.15

Table 7-2: Average value of syllable duration. Subsequently, we examined the constituents of stressed and unstressed syllables, and Table 7-3 illustrates the percentage. Vowel

Alb 50%

Chin 54%

Pol 45%

Rom 45%

Ita 54%

Table 7-3: Vowel percentage in stressed and unstressed syllables. In Chinese, the vowels in stressed syllables are longer than consonants (both in opened and closed syllables) constituting the most important part of duration, while in Polish and Romanian the opposite is the case. In unstressed syllables, Polish has the same lengthening of vowels and consonants, which depends on its complex syllabic structure. In other languages, the stressed vowel is always about 40% longer than the unstressed. Only Romanian has 66.6% of unstressed vowels, which represents its simple syllabic structure and the importance of the vocalic element. There are also important considerations which can be made regarding musical analysis.9 The analysis of prominence is very important and this is also shown by the results of a study produced on read speech, of which we here report the data in Table 7-4.

9

Romito et al., “Analisi percettiva, musicale e automatica dell’italiano L1 e L2”.

Chapter Seven

144

Languages Italian Albanian Chinese Polish Romanian

Nr prominences 4 8 2 6 6

Nr pauses 1 2 2 1 1

Table 7-4: Number of prominences and pauses. These results suggest that, for rhythm, it is important to consider other characteristics involved in languages, for example, the parameters shown by prominence, which should allow better characterization of the rhythm of natural languages. Our next research study will examine the acoustic analysis of prominence and will involve a statistical comparison. We now proceed to consideration of the data obtained from the rhythmical analysis. Figure 7-3 illustrates the PVI results.

Fig. 7-3: PVI results.

As can be noted from the Figure, there is no strict categorical distinction between stress- and syllable-timed groups for L1 and L2 Italian, but it appears that languages can be more or less syllable- or stresstimed. In this study we consider our data as a model for L1 Italian because data in relevant literature are mainly collected from spontaneous speech.

A Rhythmic-Prosodic Analysis of Italian L1 and L2

145

The two languages which manifest the greatest differences between each other are L1 Italian and L2 Italian of Chinese participants. The first has a position close to a syllable-timed group while the second is close to a stress-timed group. This result represents a linguistic transfer. Chinese speakers transfer the syllabic structure of their L1 into L2 Italian. Chinese speakers tend to dilate vocalic and consonant durations and in their first language the least unit of the language is the syllable that is composed of initial part, nucleus and final segment, therefore constituting an essential difference in the phonological system. L1 Italian has a low value for nPVI and rPVI which denotes syllable-timed languages. At the same time, we can note great variability in the production of vocalic elements, which is described by the standard deviation. In Fig. 7-3, Polish has a central position that cannot be classified in the two rhythmic groups. In this case, we can observe a strong standard deviation for vocalic and consonant elements that emphasize a great variation for these elements. Finally, Romanian and Albanian L2 Italian, typologically similar to L1 Italian, are close to that of syllable-timed language. They have low value for rPVI and a high value for nPVI, and these languages are located in the mid-low part of Fig. 7-4. Both languages present high values for standard deviation for the vocalic element. We illustrate the results of CCI in Fig. 7-4:

Fig. 7-4: CCI results.

146

Chapter Seven

For this metric, the languages located near the bisecting line are considered syllable-timed, while those distant from it are stress-timed. L2 Italian of Chinese speakers, unlike data presented for PVI, is syllable-timed and, at the same time, has a high level of standard deviation for vocalic and consonant elements. In the same way, the results obtained for L2 Italian of Polish speakers Polish L2 Italian differ from those shown in PVI. In this case, Polish speakers produce a syllable-timed L2 Italian. Both L2 Italian of Chinese and Polish speakers have results similar to that of L1 Italian described in the preview metric. Another difference that involves L2 Italian of Polish speakers compared to PVI is the strong standard deviation that regards consonant elements that prove to be more variable. L1 Italian, in this metric—and differently from the PVI results— assumes an intermediate position that cannot be considered either syllabletimed or stress-timed. Finally, L2 Italian of Romanian and Albanian speakers Romanian and Albanian L2 Italian can be considered syllable-timed, and this result concurs with that of PVI. Regardless of their classification in this metric, Romanian and Albanian speakers have a tendency to assume the same strategies of speech when speaking in L2 Italian. In Figure 7-5 we show the results for GVI with M=1.

Fig.7-5: GVI with M=1 results.

A Rhythmic-Prosodic Analysis of Italian L1 and L2

147

The interpretation of the data is the same as for PVI, because this metric is based on it. Polish, as we have seen in PVI, has a central position and cannot be classified in the two rhythmic groups. In this case, we can also observe a strong standard deviation on vocalic and consonant elements that emphasizes a great variation for those elements. This data differs from the CCI results in which it was described as a syllable-timed language. Different results compared to the previous metrics can be noted also for L1 Italian and L2 Italian of Chinese speakers Chinese L2 Italian. In this case, these languaes have a central position, different from that of L2 Italian of Polish speakers, but at the same time they cannot be classified either as syllable-timed or as stress-timed. Finally, we can note the close relationship between L2 Italian of Romanian and Albanian speakers. These languages, as can be seen from the previous metrics, can be considered as syllable-timed, and they have an adjacent position again and this confirms that the speakers use the same rhythmic characteristics when they speak in L2 Italian. This tendency is explained by the typological closeness between the languages. A consideration can be made regarding the differences between L1 Italian and L2 Italian of Romanian and Albanian speakers. As we can see from Fig.7-5, the languages mentioned above have similar or the same vocalic length, but a different consonantal duration. This result is closely related to the loss of geminate consonants by Romanian and Albanian speakers, since these are absent in their phonological system. Finally, we analyze the results for GVI with M=3 in Fig. 7-6.

Chapter Seven

148

Fig. 7-6: GVI with M=3 results.

As showed by Tepperman et al. 2011: “there we used long-distance syllable level duration and differences for improved discrimition between native and non-native English, with the best classification accuracy at M=3 context”.10

In this case, the data cannot be compared with that in the literature, because in so doing, Italian would emerge as stress-timed. We can say that all languages can be considered stress-timed or located in positions near it in the upper-right hand part of Fig. 7-6. The result for Chinese confirms the disposition of PVI, while all other languages show different preferences. In this case, L1 Italian is considered stress-timed and this data contrasts with that present in PVI and GVI with M=1 analyzed in this study. The same affirmation can be made for L2 Italian of Polish speakers and in this case we can say that this position underlines the influence of 10

Joseph Tepperman, “Testing suprasegmental English through Parroting” (paper presented at the biennal meeting for the Speech Prosody, Chicago, Illinois, May 11-14, 2010).

A Rhythmic-Prosodic Analysis of Italian L1 and L2

149

the first language that contains more complex syllabic structure and phonotactic rules. Furthermore, in these results, L2 Italian of Albanian and Romanian speakers show the same tendency and both pass from a classification of syllable-timed language to stress-timed. In these cases, the consideration of the foot shows a stress preference, which is non-syllabic, for all languages. So, all languages are located near our L1 Italian model and it is necessary to make a statistical analysis to study the tendencies and the interferences for each L1.

4. Italian Index Given the differences found in the analysis, the authors of this paper propose an index called the Italian Index which plays a dual role. On the one hand, it pursues the scientific objective of verifying the rhythmic characteristics that exist in a language, in this case the differences between L1 produced by Italian (Calabrian) speakers and L2 Italian produced by Polish, Romanian, Chinese and Albanian speakers, while on the other, it pursues an educational goal, since it identifies the critical points in speech produced by L2 Italian speakers using Italian learned without the support of any kind of school context or study aid. We report the graph of the Italian Index in Fig. 7-7.

Fig. 7-7: Italian Index.

At first glance, one can see that the language which is most distant from L1 Italian, and different also from the others, is produced by Polish and Chinese L2 Italian speakers.

150

Chapter Seven

For Polish speakers of L2 Italian, there are significant differences in the duration of vowels and consonants in stressed syllables and also in the duration of unstressed syllables and vowels. This can be linked to the syllabic structure of Polish in which the consonants tend to spread within the syllabic structure and have a longer duration than stressed vowels, while unstressed vowels and syllables are reduced to a lesser extent than stressed elements, so there is no need to reduce their length to leave space to the duration of the consonant element The consonants in stressed syllables have a longer duration than those found in unstressed syllables and the length of the stressed syllable produced by Italian speakers L1 is similar to that made by the Polish speakers of Italian L2. Therefore, the duration of the stressed syllable is compensated for through the increase of consonant duration. All of these differences are neutralized by the articulation rate, which shows the same values between L1 and L2 Italian produced by Polish speakers, and which reflects a more accurate articulatory gesture. Moreover, the languages with a complex syllable structure such as Polish tend to have low values for articulation rate, but in these cases we do not find many differences in articulation rate and this parameter will be not considered in the rest of the present study. Between L1 Italian and L2 Italian of Romanian speakers the differences are less pronounced, but still significant because they identify a different communicative intent compared to L1 Italian. The differences concern the composition of unstressed and stressed syllables. The stressed syllable is reduced, and this decrease is mainly due to the reduction in duration of the stressed vowel increasing the duration of consonants. In addition, the stressed vowel is reduced to a greater extent than the unstressed and the same process occurs for the consonants (although to a greater extent for the consonants in the stressed syllable). This trend shows that speakers maintain the difference between stressed and unstressed syllables, expanding the duration of the consonant segments and reducing vowel elements. The Romanian speakers pay more attention to the duration of the consonant segment, and this feature depends on the source language in which there exist phones which are longer than those in Italian. Although most speakers lose geminates, they recover by increasing the duration of some phones, especially the [s] which is always produced as a voiced consonant. Thus, the difference is not due to syllable structure but rather to the intrinsic duration of some consonant phones which create an increase in the consonant durations, modifying the syllabic structure of L2 Italian produced by these speakers. L2 Italian produced by Albanian speakers emerges as the closest language to L1 Italian. In effect, the differences are related only to the

A Rhythmic-Prosodic Analysis of Italian L1 and L2

151

composition of the stressed syllable. The duration of stressed vowels is reduced in the stressed syllable, but at the same time there is an increase in the consonant duration, making the total syllabic duration more similar to that of L1 Italian. It should be emphasized that the unstressed syllable and consonant do not change, nor do they maintain their durations, which are very close to those produced in L1 Italian. The speakers consider stressed vowels to be too long and tend to reduce them, increasing the duration of the consonant element. Again, the difference is not related to the syllabic structure but rather to the duration of L1 phones that modify the syllabic structure of the target language. They also transfer part of the duration of some consonants from their L1 to L2 Italian, albeit in a manner which is milder than that of the Romanian speakers, and this transfer primarily concerns the [s], which is always produced as voiced. Chinese speakers show many differences that are related to the duration of elements (vowels and consonants). This can be related to the use of duration of the elements in L1 Chinese because these speakers tend to produce, within a syllable, a longer duration of stressed and unstressed vowels, reducing the consonant duration in stressed and unstressed syllables. Stressed and unstressed syllables have a similar duration due to the vowel because, as we can see from the Figure illustrating the Italian Index, there is a strong reduction in consonant duration. Finally, we can observe a great difference between L1 Italian and other languages as regards the number and the distribution11 of prominences. We consider the prominence to be responsible for the rhythmic trend of speech and we believe that clarification of its characteristics should allow a better description of the rhythm of natural languages. We believe that these variables must be related, because it is not a single level of analysis that transmits the entire information but rather the sharing of different levels. An anomalous prosodic trend confirms ambiguous phoneticphonological productions.

5. Conclusions In conclusion, we can say that the influence of the first language on the second is evident at all levels of language.

11

Luciano Romito, Andrea Tarasi and Rosita Lio, “Italian Index: rhythmicalprosodic analysis of Italian L2 produced by Albanian, Chinese, Polish and Romanian speakers” (paper presented at the biennal conference for the V C.F.E., Càceres, Extremadura, October 25-28, 2011, in press).

152

Chapter Seven

The perceptive test and the analysis carried out on the SR system show the importance that the phonetic and phonological levels assume in speech recognition. The speakers/listeners and the SR system use the phonetic and phonological level to identify the L2 Italian. An important consideration regards intonation. In the perception test, it seems to be the case that the listeners have, as part of their competence, an intonation model which enables them to differentiate the native speaker from the non-native but also a Southern Italian speaker from one from the North. The results obtained through rhythmic metrics are conflicting in some cases. The GVI with M=3 is very different from other indices, showing a stress preference for all languages. In the other metrics, in some cases, some languages cannot be classified in the two rhythmic groups and it is important also to underline that languages have a predisposition, in the sense that they can be more or less syllable- or stress-timed. These results suggest the use of new parameters or models which enable a more specific description of the rhythm of natural languages. We believe that prominence and its parameters can be used to improve these results. This consideration is supported by the results obtained with musical analysis.12 The Italian index constitutes a significantly useful instrument in two important senses. The first is in the context of automatic speech recognition for everyday use such as in mobile communications devices, GPS systems, call centres and so on, for Italian used by non-native speakers. The second, more academically, is in its relevance both to the evaluation of rhythmic characteristics of speech supporting teacher intervention aimed at resolving the difficulties produced by L2 Italian speakers. Future research plans will involve statistical analysis and further work on prominence measurement.

12

Romito et al., “Analisi percettiva, musicale e automatica dell’italiano L1 e L2”.

CHAPTER EIGHT REQUESTING IN ITALIAN AS A FOREIGN LANGUAGE: PRAGMATIC AND PROSODIC FEATURES IOLANDA ALFANO UNIVERSITY OF SALERNO, ITALY

CLAUDIA CROCCO GHENT UNIVERSITY, BELGIUM

AND RENATA SAVY UNIVERSITY OF SALERNO, ITALY

1. Introduction This article is concerned with the pragmatic and prosodic strategies that are employed in interrogative utterances by three groups of Belgian Dutch-speaking students acquiring Italian as a Foreign Language (FL) in a classroom context. The three groups have different levels of general proficiency in Italian. The focus of the investigation is on the realization of acts of request by the three groups of students. In the first part of the paper, we examine the pragmatic strategies used by the learners to express a request for a communicative contribution. We show that requests are predominantly realized by the learners through query_y moves or yes-no questions (YNQs); in the second part of the paper we present an intonational analysis of these types of utterances. The study shows that the acquisition of pragmatic skills does not progress appreciably when the general competence in the FL increases, whereas intonation shows significant improvements in the group with a higher proficiency level. The paper is structured as follows: in section 1, we introduce the theoretical background of our pragmatic analysis. Thereafter (§ 2.), we define the speech act of requesting. In section 3, the research questions

154

Chapter Eight

addressed in this paper are outlined, together with our hypotheses on the pragmatic and prosodic strategies adopted by the learners at different stages of their acquisition process. Subsequently, we present the methodology used to collect and analyze the data (§ 4.). The subsequent subsections present and discuss the results of the pragmatic (§ 5.1.) and prosodic (§ 5.2.) analyses. Finally, we evaluate the results and draw general conclusions (§ 6.).

2. Theoretical background 2.1. Speech act theory It is widely accepted that prosody provides a systematic contribution to utterance interpretation. Interest in the contribution of prosody to pragmatic meaning has been growing in recent decades from a variety of theoretical frameworks, such as the Relevance theory,1 the Conversational analysis,2 and the Speech act theory.3 Speech act theoretical approaches to pragmatics assume that the identification of the speech act conveyed by an utterance is necessary for a successful communication.4 The correct recognition of a speech act

1

Dan Sperber and Deirdre Wilson, Relevance: Communication and Cognition (Oxford: Blackwell, 1995). 2 Emanuel A. Schegloff and Harvey Sacks, “Opening up closings”, Semiotica 8 (1973): 289-327; Harvey Sacks, Emanuel A. Schegloff and Gail Jefferson, “A simplest systematics for the organisation of turn taking for conversation”, Language 50 (1974): 696-735; Harvey Sacks, “Notes on Methodology”, in Structures of social action: Studies in conversation analysis, eds. John Maxwell Atkinson and John Heritage (Cambridge: Cambridge University Press, 1984), 21-27. 3 John L. Austin, How to do things with words (Oxford, UK: Oxford University Press, 1962); John Searle, Speech Acts (Cambridge: Cambridge University Press, 1969). 4 Among others, cf. Shoshana Blum-Kulka, Juliane House and Gabriele Kasper, Cross-cultural pragmatics: Requests and apologies (Norwood, NJ: Ablex, 1989); Anna Trosborg, Interlanguage Pragmatics: Requests, Complaints and Apologies (Berlin, New York: Mouton Gruyter, 1995); Andrew D. Cohen, “Developing the ability to perform speech acts”, Studies in Second Language Acquisition 18 (1996a): 253-67; Andrew D. Cohen, “Speech acts”, in Sociolinguistics and language teaching, eds. Sandra Lee, McKay and Nancy Hornberger (New York: Cambridge University Press, 1996b) 383-420; Kenneth R. Rose, “An exploratory cross-sectional study of interlanguage pragmatic development”, Studies in Second Language Acquisition 22 (2000): 27-67; Machiko Achiba, Learning to request in a

Requesting in Italian as a Foreign Language

155

implies the identification of the purpose(s) or communicative function(s) of an utterance in a specific situation. Among other functions, prosodic patterns lead listeners to interpret the communicative goals the speaker wishes to reach by producing a certain utterance. The association between communicative functions expressed by different speech acts and prosodic features has been explored in several studies devoted to the pragmaticsprosody interface.5 These works have demonstrated that, even if a particular syntactic and intonational structure can be preferentially employed to express particular speech acts, there is no one-to-one correspondence between linguistic (syntactic, prosodic) form and pragmatic function.6 Requests, for instance, are often realized through interrogative sentences, second language. Child interlanguage pragmatics (Clevedon: Multilingual Matters, 2003). 5 Anne Wichmann and Diane Blakemore, “The prosody-pragmatics interface,” Journal of Pragmatics 38 (2006); Dagmar Barth-Weingarten, Nicole Dehé and Anne Wichmann, Where Prosody meets Pragmatics, Studies in Pragmatics 8 (Bingley: Emerald, 2009). 6 Among others, cf. Giovanna Marotta and Patrizia Sorianello, “Question intonation in Sienese Italian”, in Proceedings of the 14th International Conference of Phonetic Sciences, eds. John J. Ohala, Yoko Hasegawa, Manjari Ohala, Daniel Granville and Ashlee C. Bailey, Berkeley: University of California, 1999: 116164; Patrizia Sorianello, “Modelli intonativi dell’interrogazione in una varietà di italiano meridionale”, Rivista Italiana di Dialettologia XXV (2001): 85-108; Martine Grice and Michelina Savino, “Intonation and communicative function in a regional variety of Italian”, Phonus 1 (1995): 19-22; Martine Grice and Michelina Savino, “Question type and information structure in Italian”, in Proceedings of Prosodic Interface, eds. Amina Mettouchi and Gaëlle Ferré (Nantes: Université de Nantes, UFR Lettres et Langage, AAI, 2003a) 117-22; Martine Grice and Michelina Savino, “Information structure and questions: evidence from taskoriented dialogues in a variety of Italian”, in Regional Variation in Intonation, eds. Peter Gilles and Jörg Peters, V/240 S. Linguistische Arbeiten (Tübingen: Niemeyer, 2004), 161-87; Martine Grice et al., “The intonation of queries and checks across languages: data from Map Task dialogs”, in Proceedings of the 13th International Congress of Phonetic Sciences 3, eds. Kjell Elenius and Peter Branderud (Stockholm: Stockholm University, 1995), 648-51; Nancy Hedberg, Juan M. Sosa and Lorna Fadden, “Meanings and Configurations of Questions in English,” in Proceedings of Speech Prosody, eds. Bernard Bel and Isabelle Marlien (Nara, Japan, 2004); Claudia Crocco, “Prosodic and informational aspects of polar questions in Neapolitan Italian”, in Proceedings of Speech Prosody, eds. Rudiger Hoffmann and Hansjorg Mixdorff (Dresden: TUCpress, 2006), 807-10; Rosa Giordano, “The intonation of polar questions in two central varieties of Italian”, in Proceedings of Speech Prosody, eds. Rudiger Hoffmann and Hansjorg Mixdorff, (Dresden: TUC press, 2006), PS8-10-155.pdf.

156

Chapter Eight

although in principle, these acts can also be expressed through other types of sentences.

2.2. Teaching pragmatics and prosody of a foreign language The knowledge required to perform speech acts in an adequate pragmatic and prosodic manner constitutes part of the communicative competence that learners of a foreign language must acquire. Pragmatics and prosody are important components of face-to-face interaction, but they are often overlooked in teaching programmes for non-native speakers. Even in a classroom context, students seldom receive explicit guidance on the acquisition of prosodic and pragmatic skills in the FL.We do not therefore know how the interlanguage7 (IL) develops from the pragmatic and prosodic point of view with respect to the global proficiency level of FL students. Many studies on L2/FL prosodic acquisition provide evidence of L1 transfer phenomena of functional prosodic patterns in L2/FL.8 Similarly, research on pragmatic transfer has underlined the impact of the learners’ L1 knowledge on their L2/FL pragmatic competences. Even if the transfer may have a positive result and not always a negative one, the crucial point is that learners seem to use pragmatic strategies conditioned by their L1.9 7

Larry Selinker, “Interlanguage”, International Review of Applied Linguistics 10 (1972): 209-41. 8 Among others, cf. Ulrike Gut, “Prosody in second language speech production: the role of the native language”, Fremdsprachen Lehren und Lernen 32 (2003): 133-52; Heather Goad and Lydia White, “Ultimate attainment in interlanguage grammars: a prosodic approach”, Second Language Research 22/3 (2006): 243-68; William Barry, “Rhythm as an L2 problem: How prosodic is it?”, in Non-Native Prosody. Phonetic Description and Teaching Practice, eds. Jürgen Trouvain and Ulrike Gut, (Berlin / New York: Mouton de Gruyter, 2007); Inneke Mennen, “Phonological and phonetic influences in non-native intonation”, in Non-Native Prosody. Phonetic Description and Teaching Practice, eds. Jürgen Trouvain and Ulrike Gut, (Berlin / New York: Mouton de Gruyter, 2007); Laurent Rasier and Philippe Hiligsmann, “Prosodic transfer. Theoretical and methodological issues”, Nouveaux cahiers de linguistique Française 28 (2007): 41-66; Marc Swerts and Sabrine Zerbian, “Prosodic transfer in Black South African English”, Proceedings of Speech Prosody, Chicago, 2010, accessed June 2012, http://speechprosody2010.illinois.edu/papers/100198.pdf. 9 Gabriele Kasper, “Pragmatic Transfer”, Second Language Research 8 (1992): 203-31; Susan M. Gass and Larry Selinker, Second Language Acquisition: An Introductory Course, 2nd edition (Mahwah, NJ: Lawrence Erlbaum Associates, 2001).

Requesting in Italian as a Foreign Language

157

Furthermore, non-native speakers perform speech acts in accord with sociolinguistic norms, varying their strategies in different contexts, but learners’ variations follow the L1 sociolinguistic patterns.10

3. Defining requests Within the framework of speech act theory, a request is “an illocutionary act whereby a speaker (requester) conveys to a hearer (requestee) that he/she wants the requestee to perform an act which is for the benefit of the speaker”.11 The requested contribution may be an action, an object, some kind of service, or a non-verbal or verbal good, i.e. information. The requests we analyze in this work belong to the second type of speech acts. Pragmatic annotation schemes, such as the Map-task Coding Scheme12 or DAMSL (dialogue act markup in several layers),13 are based upon the theoretical premises of the speech act theory. The annotation is aimed at identifying the pragmatic function(s) assumed by each speech act in its communicative context (dialogue act14). In the present study, dialogue acts 10

Patricia Bou Franch, “On Pragmatic Transfer”, Studies on English Language and Linguistics (1998): 5-18. 11 Anna Trosborg, Interlanguage Pragmatics: Requests, Complaints and Apologies (Berlin, New York: Mouton Gruyter, 1995), 187. 12 Anne Anderson et al., “The HCRC Map Task Corpus”, Language and Speech 34 (1991): 351-66; Jean Carletta et al., HCRC Dialogue Structure Coding Manual, Technical Report, 82, Human Communication Research Centre, University of Edinburgh (1996). 13 James F. Allen and Mark G. Core, DAMSL: Dialog act markup in several layers (draft 2.1), Technical report, Multiparty Discourse Group, Discourse Research Initiative, September/October (1997). 14 The same notion is indicated as “dialogue move” by Lauri Carlson, Dialogue Games: An Approach to Discourse Analysis (Dordrecht, Holland: D. Reidel Co, 1983), “dialogue object” by Arne Jönsson, “Dialogue Actions for Natural Language Interfaces”, in Proceedings of IJCAI-95, ed. Chris Mellish (Montrèal, Canada: Morgan Kaufmann Publishers, 1995), 1406-11, “communicative action” by James Allen and Mark Core DAMSL: Dialog act markup in several layers (draft 2.1), Technical report, Multiparty Discourse Group, Discourse Research Initiative, September/October 1997) and as “communicative act” by Jens Allwood, “Notes on Dialog and Cooperation”, in Proceedings of the IJCAI 97 Workshop on Collaboration, Cooperation and Conflict in Dialogue Systems, eds. Kristina Jokinen, David Sadek and David Traum (Nagoya Japan, 1997), 9–21, and David Traum, Coding Schemes for Spoken Dialogue Structure, Unpublished manuscript, 1996.

Chapter Eight

158

have been classified according to the PrATID (Pragmatic Annotation Tool for Italian Dialogues15) scheme for pragmatic annotation, which has been designed for the analysis of task-oriented dialogues elicited with the “spot the difference” technique (see below, § 3.1.). PrATID is a monodimensional scheme that does not allow for the association of multiple labels with a single dialogue act. It means that the annotation focuses of the primary pragmatic function assumed by each dialogue act in its communicative context. Such a restriction, while reducing the actual complexity of the dialogue act functions, has the advantage of avoiding the proliferation of vaguely defined sub-categories. Within the coding scheme PrATID, requests eliciting a contribution in communication can be realized by the following conversational moves.16 a)

Info_request; open question about a referent that is not yet established as a topic in the conversation: 1)

E la coda de-l cane? And the tail of-the dog? And the dog’s tail? / And what about the dog’s tail?

b) Query_w; questions introduced by an interrogative pronoun: 2)

c)

coda tail

de-l of-the

cane? dog?

Query_y; requests consisting of a YNQ: 3)

15

Com’ è la How is the How is the dog’s tail?

La

coda de-l cane è nascosta tra le zampe? The tail of-the dog is hidden between the legs? Is the dog’s tail hidden between its legs?

Renata Savy, “Pr.A.T.I.D: a coding scheme for pragmatic annotation of dialogues”, in Proceedings of LREC 2010, eds. Núria Bel, Béatrice Daile, Andrejs (Vasiljevs Malta, 2010), 2141-48. 16 Carletta et al., HCRC Dialogue Structure Coding Manual.

Requesting in Italian as a Foreign Language

159

d) Check; confirmation request: 4)

La

coda de-l cane è nascosta tra le zampe, vero? The tail of-the dog is hidden between the legs, true? The dog’s tail is hidden between its legs, isn’t it?

Examples 1-4 only represent one of the possible linguistic forms of these moves. Because there is no one-to-one correspondence between linguistic structure and pragmatic function (cf. 1.1), the requesting moves may assume different syntactic and prosodic forms. For instance, studies of Italian intonation have shown that a check move (i.e. a request for a confirmation) can be intonationally indistinguishable from a query move or can be uttered with a specific interrogative intonation pattern. This has been proven true, among others, for Bari Italian.17 In this language variety, the speakers have different tunes at their disposal for query and check moves. The choice between the tunes depends upon the attitude of the speaker toward the content of her/his utterance. When the speaker is uncertain about the information s/he wants to verify and, therefore, his confirmation request can by classified as a “tentative check”, the intonation will be indistinguishable from that of queries. In such a case, the distinction between queries and a (tentative) check is crucially a matter of contextual interpretation. In contrast, when the speaker is more confident, her/his certainty will also affect the intonation used to utter the check. In this case, the intonation profile of the check move will clearly differ from those of queries. Checks can also be explicitly introduced by discourse marker such as quindi, allora, dunque (therefore, so, then; cf. below), or can be followed by a question tag, such as sì, no, vero, eh (corresponding to “isn’t it” and similar in English; cf. below).

3.1. Studies on the acquisition of the requests The categories of requests have received considerable attention in Second Language Acquisition research, not only because they are facethreatening acts, but also because they differ cross-linguistically in several ways, so that learners must acquire what constitutes an appropriate request 17

Martine Grice and Michelina Savino, “Map Tasks in Italian: asking questions about given, accessible and new information”, Catalan Journal of Linguistics, 2 (2003b): 153-80.

160

Chapter Eight

in another language.18 Previous studies on the acquisition of requests19 have focused, among others, on the understanding of the illocutionary force of a request, on the learners’ capability to perceive the sociolinguistic meanings encoded by different kinds of requests, and on the production of appropriate requests, as they pertain to both linguistic forms and to aspects of politeness and directness. Learners appear to have no difficulties in understanding the illocutionary force of a request, probably because they rely on situational and contextual cues. They are also able to perceive sociolinguistic meanings linked with different requests, even if they may become oversensitive to them. As far as production is concerned, there is clear evidence of developmental progression: learners begin with simple and direct requests and then extend their repertoire. Longitudinal studies reported by Ellis (1994) indicate that the range of request types expanded, the range of exponents of a specific request increased and requests that encoded the hearer’s perspective (as opposed to the speaker’s) emerged. Higher proficiency should, therefore, imply a higher level of pragmatic competence.20

4. Research questions In this paper, we propose a pragmatic analysis of different moves used by non-native speakers to express a request for a contribution in communication. We subsequently describe the prosody of requests realized through query_y moves. Among the different request categories of our analysis (info_request, query_w, query_y and check), we have chosen to describe the prosody of query_y because this is the most frequent type of pragmatic move in all of the three groups of speakers we have examined. The focus of the investigation is on the comparison between learners’ and native-speakers’ performances at the pragmatic as well as at the prosodic level. Moreover, we illustrate the development of pragmatic and prosodic performances in the learners’ IL by comparing groups of learners with different proficiency levels. We will focus on the following issues:

18

Rod Ellis, The Study of Second Language Acquisition (Oxford: Oxford University Press, 1994). 19 Ellis, The Study of Second Language Acquisition. 20 Gabriele Kasper, Can pragmatic competence be taught? (1997), accessed March 2012, http://www.nflrc.hawaii.edu/NetWorks/NW06/.

Requesting in Italian as a Foreign Language

161

a. Do learners perform requests using native-like pragmatic strategies (i.e. moves)? b. Do learners realize requests using native-like morphosyntactic and prosodic structures? c. Do advanced learners perform better than beginners? We hypothesize that the beginners use pragmatic and prosodic strategies that do not correspond to those used by native speakers of Italian, but their performance improves when the general proficiency level increases. Therefore, we expect advanced learners to employ a wider range of pragmatic strategies and a more native-like intonation than do beginners. There is evidence that the acquisition of requests starts with simple and direct acts, whereas the command of complex and indirect requesting acts is acquired only subsequently.21 On the basis of these findings, we hypothesize that indirect requests (i.e. check) will be produced only by the advanced learners, whereas students of all levels will be able to perform direct requests (i.e. query_y, query_w, info_request).

5. Methodology 5.1. Corpus The corpus used for the present work is constructed out of six taskorientated dialogues elicited with the “spot the difference” technique. Each dialogue has an approximate duration of 15 minutes. The task requires the participants to find the differences between two drawings, relying only on verbal interaction.22 21

Linda L. Harlow, “Do they mean what they say? Sociopragmatic competence and second language learners”, The Modern Language Journal 74 (1990): 328–49; Caryn Francis, “Talk to me! The development of request strategies in non-native speakers of English”, Working Papers in Educational Linguistics (1997): 23–40; Montserrat Pérez i Parent, “The production of requests by Catalan learners of English: Situational and proficiency level effects”, ATLANTIS 2 (2002): 147–68; Alireza Jalilifar, “Request Strategies: Cross-Sectional Study of Iranian EFL Learners and Australian Native Speakers”, English Language Teaching 2 (2009): 46-61. 22 Vincent Péan et al., “The design and recording of icy, a corpus for the study of intraspeaker variability and the characterisation of speaking style”, in Proceedings of Eurospeech, 3rd European Conference on Speech Communication and Technology (Berlin, Germany, 1993), 27-630; Renata Savy and Francesco Cutugno, “CLIPS. Diatopic, diamesic and diaphasic variations in spoken Italian”, in On-line Proceedings 5th Corpus Linguistics Conference, eds. Michaela Mahlberg, Victorina

162

Chapter Eight

Task-orientated dialogues are “semi-spontaneous”: on the one hand, these dialogues are unplanned, without explicit limitations on the duration of the interaction or constraints on turn taking. On the other hand, the interaction is curtailed by the assignment, and pragmatic, syntactic and lexical variation is reduced because of the predominance of the question/answer structure in the text and because lexical choices are constrained by the task design (e.g. the presence of certain objects in the drawings). The requests occurring in the six dialogues have been classified according to the PrATID scheme (§ 2.) as info_request, query_w, query_y and check. In this type of dialogue, requests are used to gather information about the details of the picture and therefore, they play a central role in the accomplishment of the task.

5.2. Participants and procedure The 12 speakers were female native speakers of Belgian Dutch who were students of Italian at Ghent University, with ages ranging from 2025. At the time of the recordings, the speakers attended courses of Italian as a FL, which ideally provide the students with a competence level in Italian corresponding to the CEFR levels B1, B2 and C1.23 We recorded 4 speakers for each competence level. The learners had never visited Italy for longer than two weeks and had not taken any Italian classes prior to their university studies. The dialogues were recorded with two headset microphones (Shure WH20QTR) plugged into an M-Audio Microtrack 24/96. The recordings were made in a quiet room, with only one of the authors and the two speakers present. The speakers were not aware of the goal of the recordings and were reassured that the task was not meant to measure their IQ or their language proficiency. They were permitted to speak only Italian, and received simple instructions to follow, including a short description of the task (i.e. discovering seven slight differences between the pictures without looking at each other). After a few minutes, the students seemed to be at ease, and their interactions became increasingly spontaneous.

González Díaz and Catherine Smith, University of Liverpool, UK, 2009, accessed June 2012, http://ucrel.lancs.ac.uk/publications/cl2009/, article 213. 23 Common European Framework of Reference for Languages, accessed March 2012, http://www.coe.int/t/dg4/linguistic/CADRE_EN.asp.

Requesting in Italian as a Foreign Language

163

6. Results 6.1. Pragmatic analysis The analysis was carried out on a sample of 104 request acts taken from the six spot the difference dialogues. The utterances are divided as follows among the three groups of learners. Table 8-1 illustrates in each row the number of occurrences obtained for each level of competence (B1, B2, C1) and in each column the number of occurrences of the kinds of requests analyzed (info_request, query_w, query_y, check). Note that the speakers with a higher proficiency level in the FL (C1) produced fewer request acts. The results of the analysis are graphically summarized in the bar charts presented in Fig. 8-1. B1 B2 C1

IR 3 10 0

QW 4 12 2

QY 21 33 12

CK 5 2 0

Tot. 33 57 14

Table 8-1: Count of the different kinds of requests examined. B1, B2, C1 = levels of competence; IR = info_request; QW = query_w; QY = query_y; CK = check. As shown, the different request moves are not equally represented in the sample. The majority of the requests (64%) are realized by query_y moves, independently of the speaker’s level of competence in Italian. In contrast, the other types of request moves, i.e. query_w, info_request and check moves are far less frequent in the production of the learners. This suggests that a higher proficiency level in Italian as a FL does not imply the use of indirect and more complex requests: checks are 7% of the requests, and do not appear in the productions of the highest proficiency level group. This result shows that all three groups of speakers prefer more direct and simple requests. The requesting strategies put forward by the learners differ appreciably from those employed by native speakers of Italian. Firstly, non-native speakers, independently of their proficiency level in the FL, perform a smaller number of requests as compared with native speakers.

164

Chapter Eight

Fig. 8-1: Types of requests (%). B1, B2, C1 = competence levels; IR = info_request; QW = query_w; QY = query_y; CK = check.

Requesting in Italian as a Foreign Language

165

A comparison between the interactions analyzed for this study and a sample of task-orientated dialogues of the same duration produced by natives24 has indicated that non-native speakers perform an average of 35 requests per dialogue in the FL, whereas in native dialogues the average is 84 requests per dialogue. Therefore, the learners produce 58% fewer requests than do native speakers. Moreover, learners with a higher proficiency level (C1) produce the fewest requests (14% of the sample): these speakers prefer describing their picture rather than asking questions of the listener. As far as morphosyntactic structures are concerned, all groups of learners mostly use a few, syntactically simple structures (89% of the occurrences): 5) Presentatives, there is/are sentences (29% of the occurrences) Ci sono piccole finestre? There are small windows? Are there small windows? 6) Subject-Verb-Object (29%) E l’ uomo ha la televisione ne-lla mano destra? And the man has the television in-the hand right? And the man has the television in his right hand? 7) Verb-2S-Object (31%) Vedi tutta la See.2S all the Can you see the whole car?

macchina? car?

Moreover, a number of morphosyntactic features that are typical of spoken interactions in Italian25 do not occur in the learners’ dialogues. For 24

The sample contains 6 spot-the difference dialogues from the CLIPS corpus, collected in Rome, Naples, Lecce, Bari, Catanzaro and Palermo. See Renata Savy and Marina Castagneto, “Funzioni comunicative e categorie d’analisi pragmatica: dal testo dialogico allo schema xml e viceversa”, in Linguistica e modelli tecnologici di ricerca, Atti del XL Congresso SLI, ed. Giacomo Ferrari (Roma: Bulzoni, 2006), 569-79. 25 Rosanna Sornicola, Sul parlato (Bologna: Il Mulino, 1981); Francesco Sabatini, “L’italiano dell’uso medio: una realtà tra le varietà linguistiche italiane”, in

Chapter Eight

166

instance, students do not utilize syntactic structures with marked word orders, such as dislocation, hanging topic or clefting, nor produce the pronominal form of avere (to have), averci (there have): these features are by contrast usual for natives.26 The following utterance (8), which is taken from an Italian spot the difference dialogue (CLIPS corpus), is a typical example of a request produced by a native. In this sentence, the pronominal verbal form averci occurs in a clitic right-dislocated construction: 8) Il

cavaliere

ce

l’

ha

la

it

has

the

spada? The

knight there sword? Does the knight have it, the sword?

In 25% of the analyzed utterances, the learners end their requests by adding a disjunctive question tag, namely o no? (or not?) or, simply, o?: 9) Il

ragazzo

tiene

l’

anatra o

the

duck

no? The

boy holds not? Does the boy hold the duck or not? 10) Tu

ved-i i due mani ragazzo o? You see-2S the two hands boy or? Do you see the two hands of the boy or...?

or

de-l of-the

This structure functions as a phatic signal aimed at maintaining communication with the interlocutor, as well as an explicit mark of the interrogative modality. Because in Italian the interrogative modality is expressed only by prosodic features, the addition of a question tag insures the transmission of the right communicative function.

Gesprochenes Italienisch in Geschichte und Gegenwart, eds. Günter Holthus and Edgar Radtke (Tübingen: Narr, 1985), 154-84. 26 Massimo Aureli et al., “Aspetti morfosintattici”, in Italiano parlato. Analisi di un dialogo, eds. Federico Albano Leoni and Rosa Giordano (Napoli: Liguori, 2005), 119-35.

Requesting in Italian as a Foreign Language

167

Furthermore, learners use question tags and discourse markers in a non-native fashion, consequently producing requesting structures without direct correspondence in Italian L1. Discourse markers, such as quindi, allora, dunque (therefore, so, then) are used by native Italian speakers to introduce a request of confirmation (check). In contrast, in non-native speech, the same markers are employed as discourse fillers rather than to introduce a check move. In addition, checks realized by native speakers of Italian often end with a question tag, such as sì?, no?, vero? or eh? (cf. ex. 4), clearly indicating that the speaker is almost sure of the information he/she is checking. Instead, rather than marking the degree of certainty of the speaker who is asking for confirmation, in non-native speech, the question tags are used to reinforce the expression of the interrogative modality in query_y moves. Given the importance of prosody for the expression of interrogativity in Italian, we decided to examine the intonational realization of the query_y moves occurring in our corpus. Because in query_y moves that are not followed by a question tag, prosody is the only cue to the expression of the intended communicative function, we have focused on these cases. The results of the prosodic analysis are presented in the following section.

6.2. Prosodic analysis 6.2.1. Measurements and results In what follows, we provide a prosodic analysis of the 45 query_y moves without question tags that occur in our corpus. These moves are prosodically realized as YNQs. The results of the analysis show that the learners do not realize the interrogative pattern in a native-like way: they produce a sharp rise that begins on the stressed vowel, which does not appear in the input provided to the students by the Italian-speaking lecturers.27 However, the results also suggest that the realization of the question rise changes in the advanced students (level C1) compared with the beginners (levels B1 and B2). The 45 query_y moves/YNQs are divided as follows among the 3 proficiency groups: - B1: 15 utterances 27

The rise seems to be a feature of the yes-no questions in the learners’ L1 (Belgian Dutch). For an analysis of the prosodic transfer in the of the yes-no questions produced by the same learners, see Baele, this volume, chapter nine.

168

Chapter Eight

- B2: 19 utterances - C1: 11 utterances YNQs are predominantly realized with a low turning point (L) corresponding to the last lexical stress of the utterance. The position of the turning point may vary: it can occur before the onset of the stressed syllable as well as later, e.g. in the middle of the syllable. After the L turning point, a steep pitch rise begins and continues until the end of the utterance, at which point it reaches the highest pitch values of the utterance (H). This pattern LH is schematically represented in Fig. 8-2. Note that this pattern matches a typical YNQ profile used in Belgian Dutch.28

Fig. 8-2: Question rise produced by the Dutch-speaking learners. The last lexically stressed syllable of the utterance is in grey.

The analysis focuses on two phonetic features, i.e. the position of the turning point L (“elbow”) relative to the onset of the stressed vowel and the pitch excursion between L and the final H. It is well known that the individuation of a low turning point is problematic because the transitions between levels and the rising/falling movements of F0 and are often continuous. As demonstrated by Del Giudice et al.,29 the performances of algorithms for automatic elbow detection are not free of errors and, furthermore, manual annotation is less inconsistent than has been said. Therefore, we identified the elbows by relying on visual judgment. We labelled the turning point as the correspondence of a local minimum after which the pitch constantly rises. The latency of the elbow has been measured relative to the stressed vowel onset (onV). The results show that the latency values (L to onV) vary from group B1 (mean latency value = 35 ms) to group B2 (mlv = 43 ms) and group C1

28

Ibidem. Alex Del Giudice et al., “Comparing methods for locations pitch elbows”, in Proceedings of 16th ICPhS, eds. Jürgen Trouvain and William Barry (Saarbrücken: Universität des Saarlandes, 2007), 1117-20. 29

Requesting in Italian as a Foreign Language

169

(mlv = 65 ms). The median latency values for the 3 groups are compared in the boxplots in Fig. 8-3. The results of a t-test show that there is no significant difference between the mean latency between group B2 and B1and between group B2 and C1. However, the difference becomes significant when group B1 and B2 are pooled and their mean latency value is compared with that of C1 (p = 0.0069; see Fig. 8-4). Therefore, the L turning point occurs later in the advanced speakers compared with the beginners. The result also suggests that the L displacement takes place gradually.

Fig. 8-3: Median latency values of the turning point L (L to onV).

Fig. 8-4: Median latency values L to onV, pooling B1 and B2 (right boxplot) vs. C1 (left boxplot).

170

Chapter Eight

We also measured the frequency excursion between the elbow L and the highest point reached by the subsequent rising. The measurements show that the mean excursion values do not differ significantly from one proficiency group to another: B1 = 92 Hz / 6,3 St; B2 = 93 Hz / 6,5 St; C1 = 92 Hz / 6,6 St. The results are graphically represented in the boxplot shown in Fig. 8-5.

Fig. 8-5: Median latency values of the of the f0 excursion between L and H in the 3 competence levels.

To summarize, whereas the frequency excursion between the starting and the end point of the pitch rise is stable in the different competence groups we have examined, the point at which the pitch rise begins (i.e. the L turning point or “elbow”) varies significantly from the beginner groups (levels B1 and B2) to that of the advanced learners (C1). In what follows, we examine the prosody of the YNQs in the varieties of Italian that form the linguistic input offered to the learners by the lecturers. We argue that the displacement of the L turning point reflects the learners’ effort to imitate the intonation of the YNQs produced by the lecturers. We also claim that the learners’ performance improves when the overall proficiency level in the FL increases. 6.2.2. Input provided to the students The learners that have been examined for the present study are exposed to different Italian accents. It is well known that Italian intonation varies

Requesting in Italian as a Foreign Language

171

considerably across the country.30 Moreover, the use of standard pronunciation requires a specific training that is not provided at schools or universities. Therefore, only professional speakers and actors speak with a standard accent, whereas other speakers keep a more or less regional pronunciation, even when they are more highly educated. At the university where the students studied, the Italian courses are taught by two native Italian speakers and by two near-native lecturers who have Belgian Dutch as L1. The native speakers are from Naples (Southern variety) and Varese (near Milan; Northern variety). One of the two near-native teachers acquired Italian with a Northern accent (Turin and Saluzzo, Piedmont) and the other has no recognizable regional accent. In the varieties of Italian that form the input offered to the students, the pitch rise characterizes the YNQs far less than it does in Belgian Dutch (cf. above). The rise is present only in Milanese Italian (YNQ tune: H+L* L- H%),31 whereas it can be absent in the variety spoken in Turin (H* or L+H* L- H%; L*+H H- L%)32 and it does not appear normally in Neapolitan Italian (L*+H HL- L%).33 Figures 6, 7 and 8 show examples of YNQs in Milanese, Turinese and Neapolitan Italian.34 Note that, when the rise is present, it begins after the stress syllable (cf. Figures 8-6, 8-7, 8-8). 6.2.3. Approaching Italian intonation The data presented above and those provided by the comparison between the productions in Italian (FL) with those in Belgian Dutch (L1)35 suggest that YNQs in Belgian Dutch and Italian differ in an important feature: in Belgian Dutch, the question rise begins early (i.e. before the

30

Barbara Gili Fivela et al., “Varieties of Italian and their ToBI transcription”, Paper presented at the Workshop on Romance ToBI (PaPI 2011), Terragona, June 23, 2011; Martine Grice et al., “Strategies for Intonation Labelling across Varieties of Italian”, in Prosodic Typology ed. Jun Sun-A (Oxford: Oxford University Press, 2005), 362-90. 31 Gili Fivela et al., “Varieties of Italian and their ToBI transcription”. 32 Gili Fivela, ibidem; Grazia M. Interlandi, “Il continuum della variazione pragmatico-espressiva nell’intonazione dell’italiano parlato a Torino”, in Il parlato italiano, Atti del Convegno nazionale, eds. F. Albano Leoni et al., (Napoli: M. D’Auria Editore, CD-Rom, 2004), 1-20. 33 Gili Fivela, “Varieties of Italian and their ToBI transcription”. 34 The utterances presented in Figures 8-6, 8-7 and 8-8 are taken from taskorientated dialogues collected in CLIPS, Savy and Cutugno, “CLIPS. Diatopic, diamesic and diaphasic variations in spoken Italian”. 35 Baele, this volume.

172

Chapter Eight

onset of the stressed syllable or by the middle of the stressed vowel), whereas in Italian, the rise (if present) begins after the stressed syllable. On the basis of these observations, we argue that the progressive displacement of the L turning point corresponds to an improvement of the prosodic competence. We assume that the learners try to imitate the model provided by their Italian-speaking lecturers and, therefore, that the learners’ productions improve when they approximate the native speakers’ performance. Belgian Dutch YNQs with an early rise do not have a counterpart in Italian: the question rise begins later in Italian. Therefore, the displacement of L, i.e. a later rise, is necessary to produce a more native-like Italian intonation. The data show that the beginners (groups B1 and B2) transfer the early pitch rise of Belgian Dutch to the FL, whereas the rise tends to begin later in the utterances of students with a higher competence level. We therefore claim that the displacement of L by the advanced learners reflects an improvement in the intonational competence because these speakers replace a feature (i.e. the early rise) that is incompatible with the input provided by the lecturers with a feature that makes their question intonation closer to that used in Italian.

Fig. 8-6: YNQ (E ci son tre tasti bianchi? And are there three white keys?) in Turinese Italian: L*+H H- L%.

Requesting in Italian as a Foreign Language

173

Fig. 8-7: YNQ (Ma in trasversale? But in transverse?) in Neapolitan Italian: L*+H H- L%.

Fig. 8-8: YNQ (Tu lo vedi il dado? Do you see the dice?) in Milanese Italian: H+L* L-H%.

174

Chapter Eight

7. Conclusions On the basis of the results presented in the previous sections, we will now attempt to answer our three research questions: a. Do learners perform requests using native-like pragmatic strategies (i.e. moves)? b. Do learners realize requests using native-like morphosyntactic and prosodic structures? c. Do advanced learners perform better than beginners? Before drawing a general conclusion to this study, it is necessary to stress that the results must be verified on a larger corpus, and on other types of speech. The number of requests realized by the learners in the spot the difference dialogues used for this study is small and therefore only allows provisional conclusions. Our data show that learners perform requests using pragmatic strategies that do not correspond to those used by native speakers: in nonnative interactions, query_y moves predominate over all other types of requests, whereas native speakers performing task-oriented dialogues use a much wider range of requesting moves. Moreover, advanced learners (C1) produce fewer requests than do the students at the other two levels. This unexpected result can be tentatively attributed to the increase of the general proficiency level of the students, who prefer showing their linguistic ability by producing relatively long turns rather than by requesting information from the conversational partner. Furthermore, in all three competence levels, the requests are syntactically simple, being realized mostly as presentative or SVO structures. In addition, a number of typical features of spoken Italian (such as sentences with marked word order or pronominal forms of verbs like avere, to have) that occur in native productions do not occur in the analyzed corpus. It is worth noting that these features are usual in speech, whereas they are uncommon and even stigmatized in written language. We can therefore explain the absence of morphosyntactic features of spoken language in the learners’ productions by the classroom-learning context in which students acquire Italian. Finally, students of all competence levels use question tags and discourse markers differently from natives: learners employ question tags (o? o no?) primarily to express interrogativity unambiguously, whereas discourse markers are used as fillers. In summary, our results show that non-natives use a subset of the syntactic and pragmatic structures available to natives. The predominance

Requesting in Italian as a Foreign Language

175

of query_y moves confirms the hypothesis that learners begin by using direct and less complex requests (query_y).36 However, in opposition to what has been supposed, a higher proficiency level in Italian as an FL does not imply the use of indirect and more complex requests (check moves). Indeed, all three groups of speakers prefer to use the more direct and simple requests. Moreover, comparing groups B1 and B2, we can observe that the latter perform more requests, whereas learners with the higher proficiency level (C1) adopt a different strategy to solve the task, as they describe their own picture rather than asking questions of the interlocutor. This apparent difference in the pragmatic strategies used by the speakers of group C1 vs. those of groups B1 and B2 represents a problematic result, which, in our opinion, needs further investigation to be confirmed and interpreted. The results of the prosodic analysis show that the prosodic profile of YNQs (query_y moves) produced by the beginning learners is not nativelike: this conclusion arises from the comparison of the learners’ productions with the question tunes that occur in the regional varieties spoken by the lecturers. Given the apparent similarity between the YNQ pattern produced by the beginning learners in their mother tongue and in Italian, we argue37 that in the first stage of the interlanguage development there is a prosodic transfer from L1 to FL. However, the learners’ performance improves when the global competence level increases. This claim is supported by the results showing the differences in the latency values of the turning point at which the question rise begins: in the beginners’ productions (groups B1 and B2), the rise begins around the onset of the accented vowel, whereas advanced speakers (C1) tend to begin the rise later. We have interpreted the displacement of the turning point L as an improvement in the prosodic competence of the student with a higher global proficiency level because the YNQ with a later rise represents a better approximation of the natives’ performance. In conclusion, request acts show that prosodic competence progresses in the group of the advanced learners, whereas pragmatic skills do not improve in a clear way from one proficiency group to another as far as the range of requesting moves is concerned. An unexpected result is the advanced use of a different pragmatic strategy to accomplish the task, which is based upon explaining rather than requesting. This result raises questions about how pragmatic skills develop in the interlanguage of advanced learners. We leave this issue open for future research. 36 37

Ellis, The Study of Second Language Acquisition. Following the analysis presented in Baele, this volume.

CHAPTER NINE YES-NO QUESTIONS IN BELGIAN DUTCH (L1) AND ITALIAN (FL) JENS BAELE GHENT UNIVERSITY, BELGIUM

1. Introduction In recent years, the acquisition of second language (L2) or foreign language (FL) intonation has drawn increasing attention from researchers.1 Most of the research in the field of intonational acquisition has been performed in a country where the second language is spoken.2 Therefore, 1

See a.o. Motoko Ueyama, “The phonology and phonetics of second language intonation: the case of Japanese English”, in Proceedings of Eurospeech ‘97, eds. George Kokkinakis, Nikos Fakotakis and Evangelos Dermatas (Rhodes: University of Patras, 1997), 2411-14; Motoko Ueyama and Sun-Ah Jun, “Focus realization in Japanese English and Korean English intonation”, in Japanese/Korean Linguistics, eds. Noriko Akatsuka, Hoji Hajime and Shoichi Iwasaki (Stanford: Center for the Study of Language and Information, 1998), 7: 629-45; Ineke Mennen, “Second Language Acquisition of Intonation: The Case of Peak Alignment”, in CLS 34-2: The Panels, eds. M. Catherine Gruber et al. (Chicago: Chicago Linguistic Society, 1998), 327-41; Ineke Mennen, “The realization of nucleus placement in second language intonation”, in Proceedings of ICPhS ‘99, ed. John J. Ohala (San Francisco: University of California, 1999), 1: 555-58; Ineke Mennen, “Bidirectional interference in the intonation of Dutch speakers of Greek”, Journal of Phonetics 32/4 (2004): 543-63; Sun-Ah Jun and Mira Oh, “Acquisition of Second Language Intonation”, in Proceedings of ICSLP 2000, eds. Baozong Yuan, Taiyi Huang and Xiaofang Tang (Beijing: China Military Friendship Publishers, 2000), 4: 76-79; Jürgen Trouvain and Ulrike Gut, eds., Non-native prosody (Berlin: Mouton de Gruyter, 2007). 2 Motoko Ueyama, “The phonology and phonetics of second language intonation”, Japanese (L1) and English (L2); Motoko Ueyama and Sun-Ah Jun, “Focus realization in Japanese English and Korean English intonation”, Japanese/Korean

178

Chapter Nine

these studies concern the acquisition of an L2, i.e. a language acquired in a submersion/immersion context by learners who live in the country where the language has a primary communicative function. Other studies focus on the intonational acquisition of an FL, i.e. a language that is acquired in a classroom setting and that has no specific function in the country where it is learned.3 An important aspect of the research on intonational acquisition concerns the interference of L1, i.e. the transfer of prosodic features from L1 to L2/FL during the acquisition process. This paper addresses the interrogative intonation produced by a group of Belgian Dutch-speaking students who acquire Italian in a classroom context at university. The research focuses on the intonation of yes-no questions and presents a comparison between the productions of the speakers in their FL (Italian) and in their L1 (Belgian Dutch). The analysis focuses on a number of intonational features observable in Belgian Dutch yes-no questions and provides preliminary evidence that these features are transferred from the mother tongue to the FL. The intonational analysis presented in this paper is based on the results of previous research on Dutch intonation carried out in the framework of the autosegmental-metrical (AM) theory.4 A fundamental aspect of the AM theory is the separation of the phonological form from the phonetic realization of the melodic events (pitch accents and edge tones). The application of this tenet to the study of intonational acquisition and prosodic transfer enables the identification of the actual source of the transfer.5 Interference on the phonological level can cause differences between the tonal inventories of native speakers and of learners of a certain language (e.g. a high pitch accent H* of L1 used instead of the low (L1) and English (L2); Sun-Ah Jun and Mira Oh, “Acquisition of Second Language Intonation”, English (L1) and Korean (L2). 3 See, for example, Laurent Rasier, “Prosodie en vreemdetaalverwerving. Accentdistributie in het Frans en het Nederlands als vreemde taal” (PhD diss., Université catholique de Louvain, 2006). 4 John A. Goldsmith, Autosegmental Phonology, PhD diss., Massachusetts Institute of Technology, 1976 (New York: Garland, 1979); Janet B. Pierrehumbert, The phonology and phonetics of English intonation, PhD diss., Massachusetts Institute of Technology, 1980 (Bloomington: Indiana University Liberian Collections, 1988); D. Robert Ladd, Intonational phonology (Cambridge: Cambridge University Press, 1996); D. Robert Ladd, Intonational phonology, 2nd ed. (Cambridge: Cambridge University Press, 2008). 5 Mennen, “Bi-directional interference”; Ineke Mennen, “Phonological and phonetic influences in non-native intonation”, in Non-native prosody: Phonetic Description and Teaching Practice, eds. Jürgen Trouvain and Ulrike Gut (Berlin: Mouton de Gruyter, 2007), 53-76.

Yes-No Questions in Belgian Dutch (L1) and Italian (FL)

179

pitch accent L* of the target language), whereas interference on the phonetic level can explain differences between native and non-native speakers in the concrete realization of the same phonological tonal event (e.g. a bitonal low-high pitch accent L*+H realized with an early peak by the learners, while in the target language, there is a late peak).6 The comparison between the learners’ productions in FL (Italian) and in L1 (Belgian Dutch) presented in this paper suggests that the learners transferred the interrogative tunes of their mother tongue to the FL, Italian. The data provided by Iolanda Alfano, Claudia Crocco and Renata Savy in this volume, however, provide evidence that the transfer of intonational features affects the productions of beginners and advanced learners to different degrees.7 The paper is structured as follows: section 1 briefly presents the current intonational analyses of yes-no question intonation in Dutch. The corpus and the methodology used for this study are introduced in section 2. Thereafter, (§ 3) the results of the analysis of the Belgian Dutch corpus (§ 3.1) and of Italian as FL (§ 3.2) are presented. In subsection 3.3, the data from the two corpora are compared. The conclusions are drawn in section 4.

2. Yes-No question intonation in Dutch The intonation of Dutch has been studied intensively, especially from an AM perspective.8 Studies on the prosodic pattern of yes-no questions in 6

Ineke Mennen, “Phonological and phonetic influences in non-native intonation”, 57. 7 See chapter eight. 8 See a.o. Carlos Gussenhoven, “Adequacy in intonation analysis: The case of Dutch”, in Autosegmental Studies on Pitch Accent, eds. Harry van der Hulst and Norval Smith (Dordrecht: Foris, 1988), 95-122; Carlos Gussenhoven, “Tone Segments in the Intonation of Dutch”, in The Berkeley Conference on Dutch Linguistics 1989, eds. Thomas F. Shannon and Johan P. Snapper (Lanham: University Press of America, 1991), 139-55; Carlos Gussenhoven and Toni Rietveld, “A target-interpolation model for the intonation of Dutch”, in Proceedings of ICSLP ‘92, eds. John J. Ohala et al. (Banff: University of Alberta, 1992), 1235-38; Rob van den Berg, Carlos Gussenhoven and Toni Rietveld, “Downstep in Dutch: Implications for a model”, in Papers in laboratory phonology II: Gesture, Segment, Prosody, eds. Gerard J. Docherty and D. Robert Ladd (Cambridge: Cambridge University Press, 1992), 335-59. Studies on Dutch intonation have also been performed in the framework developed at the IPO (Instituut voor Perceptie Onderzoek, Eindhoven, the Netherlands), cf. René Collier and Johan ’t Hart, Cursus Nederlandse Intonatie (Leuven: Acco, 1981); René Collier, “On the

180

Chapter Nine

Dutch have shown that this intonation contour is consistently defined by a pitch rise at the end of the utterance.9 Moreover, Caspers (1998) has proved this rising movement to be indispensable for a yes-no question to be perceived as such. An extensive overview of the prosodic properties of the interrogative modality in Dutch is found in the work of Haan (2002). According to this study, yes-no questions in Dutch are predominantly realized with an H*L H% tune. In addition, the author reports three other, rather marginal, contours: H* H%, L*H H% and L* H%.10 The tonal descriptions proposed by Haan follow the principles of ToDI (Transcription of Dutch Intonation),11 a transcription system designed for Standard Dutch that is similar to ToBI12 and based on the AM framework. Contrary to ToBI, however, phrase accents are not used in ToDI, and postnuclear F0 movements preceding the boundary tone are analyzed as trailing tones added to the nuclear pitch accent. This analysis, therefore, allows for the presence of tritonal pitch accents, such as L*HL H%. This tune would be transcribed in ToBI as L*+H L-H%. Note that the studies mentioned above report findings concerning the intonation of Standard Dutch spoken in the Netherlands, whereas the L1 of the learners examined for the current research is Belgian Standard Dutch, which is spoken in Flanders (Belgium).13 The Belgian variety of Dutch has phonology of Dutch intonation”, in Worlds behind words: Essays in Honour of Professor F.G. Droste on the Occasion of his Sixtieth Birthday, eds. Frans J. Heyvaert and Frieda Steurs (Leuven: Leuven University Press, 1989), 245-58; Johan ’t Hart, René Collier and Antonie Cohen, A perceptual study of intonation. An experimental-phonetic approach to speech melody (Cambridge: Cambridge University Press, 1990). 9 See a.o. Charlotte Gooskens and Vincent J. van Heuven, “Declination in Dutch and Danish: Global versus local pitch movements in the perceptual characterisation of sentence types”, in Proceedings of ICPhS ‘95, eds. Kjell Elenius and Peter Branderud (Stockholm: Stockholm University, 1995), 2: 374-77; Vincent J. van Heuven and Judith Haan, “Phonetic correlates of statement versus question intonation in Dutch”, in Intonation. Analysis, Modelling and Technology, ed. Antonis Botinis (Dordrecht: Kluwer Academic Publishers, 2000), 119-43. 10 Judith Haan, Speaking of questions: an exploration of Dutch question intonation (Utrecht: Landelijke Onderzoekschool Taalwetenschap, 2002), 111-13. 11 Carlos Gussenhoven, “Transcription of Dutch intonation”, in Prosodic Typology: The phonology of intonation and phrasing, ed. Sun-Ah Jun (Oxford: Oxford University Press, 2005), 118-45. 12 Mary E. Beckman and Gayle M. Ayers, “Guidelines for ToBI Labelling. Version 3.0” (unpublished manuscript, March 1997), pdf file. 13 The Dutch-speaking region covers a continuous area divided between two different countries: the northern part forms the Netherlands, while the southern part forms Flanders (Belgium). The same language is spoken on both sides of the

Yes-No Questions in Belgian Dutch (L1) and Italian (FL)

181

received limited attention in literature about intonation. The studies on this variety mostly concern Limburg dialects.14 Belgian Dutch has further been analyzed by Laurent Rasier, who investigated focus intonation.15 Consequently, the existing literature provides limited evidence with regard to studying interference. However, as we will show, the intonational analysis of yes-no questions proposed for Standard Dutch spoken in the Netherlands16 can be extended to Belgian Dutch.

3. Corpus and Methodology A group of 12 university students between 19 and 21 years old who were studying Italian at Ghent University participated in this research. All of them were female and monolingual native speakers of Belgian Dutch.17 Out of these 12 speakers, 10 came from the province of East Flanders, and the remaining two came from West Flanders and Flemish Brabant. At the time of the recording, the speakers were attending Italian courses during the second, third or fourth years of their university education. The speakers were divided into three groups of equal size according to their proficiency levels in the FL, which were defined based on the linguistic competences that should ideally be obtained at the end of their respective courses. Within the CEFR (Common European Framework of Reference for Languages) the second, third and fourth years correspond to the levels B1, B2 and C1, respectively, which means that the target proficiency level of the Italian courses attended by the students ranged from intermediate to near-native. At the start of their first year, the students had all received a week long pronunciation training; however, the training did not deal explicitly with intonation.18 All recordings were made in a quiet room

border, with differences in the pronunciation of sounds and, to a lesser degree, lexicon and syntax. On the basis of these differences, a northern (i.e. Netherlandic Dutch) and a southern (i.e. Belgian Dutch) variety of Dutch can be identified. 14 Jörg Peters, “A bitonal lexical pitch accent in the Limburgian dialect of Borgloon”, in Tones and Tunes. Volume 1: Typological Studies in Word and Sentence Prosody, eds. Tomas Riad and Carlos Gussenhoven (Berlin: Mouton de Gruyter, 2007), 167-98; Jörg Peters, “Tone and intonation in the dialect of Hasselt”, Linguistics 46/5 (2008): 983-1018. 15 Rasier, “Prosodie en vreemdetaalverwerving”. 16 Haan, Speaking of questions; Gussenhoven, “Transcription of Dutch intonation”. 17 “Monolingual” means having only one mother tongue. 18 Note that the corpus of Italian FL used for this study is the same used by Alfano, Crocco and Savy, this volume, chapter eight.

182

Chapter Nine

provided by the Linguistics Department of Ghent University. Two different corpora were collected, one for FL and one for L1.

3.1. Italian (FL) corpus The productions in FL were elicited using the “spot-the difference” technique.19 In “spot-the difference” dialogues, a couple of speakers have to identify a number of differences between two versions of the same drawing. There is no eye contact between the speakers, so they must rely on verbal interactions to accomplish the task. In spite of the restrictions imposed by the task and the recording context, dialogues elicited with this technique are fairly spontaneous. yes-no questions that were perceptually identifiable as such were drawn out of the recorded dialogues. The data of two of the learners were not used because these learners did not produce yes-no questions during their dialogue. Consequently, the corpus of yes-no questions in FL is comprised of the productions of 3 learners for the first proficiency level, 4 for the second and 3 for the highest competence level. 3.1.1. Types of yes-no questions in the Italian FL corpus The data consist of 61 yes-no questions subdivided into 2 groups according to the pragmatic functions the questions served in the dialogues. The pragmatic classifications of the yes-no questions are based on the annotation scheme of conversational moves elaborated for the Map Task dialogues.20 According to this scheme, yes-no questions can realize a “query” or a “check” move.21 Queries are conversational moves requesting new information, whereas checks request confirmation of a piece of information that the speaker already knows but of which he is not entirely sure. In the latter case, the speaker has an expectation regarding the answer 19

Vincent Péan, Sheila Williams and Maxine Eskenazi, “The design and recording of icy, a corpus for the study of intraspeaker variability and the characterisation of speaking styles”, in EUROSPEECH 1993 (Berlin: European Speech Communication Association, 1993), 627-30. 20 Jean Carletta et al., “HCRC Dialogue Structure Coding Manual” (unpublished manuscript, June 3, 1996), pdf file. 21 Note that the pragmatic classification adopted in this work may not coincide with the one used on the same corpus by Alfano, Crocco and Savy, this volume, chapter eight. The classification provided by the latter is based on the PraTID scheme. The map task and the PraTID scheme differ primarily with regard to the identification of checks.

Yes-No Questions in Belgian Dutch (L1) and Italian (FL)

183

to the question. This pragmatic distinction has been adopted because there is evidence that differences on the pragmatic level may correspond to differences on the prosodic level.22 Yes-no questions expressing queries can be either disjunctive or nondisjunctive. Disjunctive questions are characterized by the tag “o no?”, which is sometimes not completed by the speaker, who produced only, “o?”. A third tag used by the learners is “o qualcosa?”, which occurred only two times: (1)

Ved-i l’ ombelico See-2S the navel Do you see the navel or not?

o or

no? not

(2)

Su-lla statua c’ è qualcosa On-the statue there is something On the statue is there something written or?

scritto o? written or

(3)

Sul-la casa ved-i un un un-a qualcosa? On-the house see-2S a-M a-M a-F something On the house do you see a door or something?

porta

o

door

or

Contrary to queries, checks do not display differences with regard to their morphosyntactic structure. All checks consist of a single phrase, which is usually nominal. A subdivision of this group was therefore unnecessary (see ex. 4). (4)

Uno specchio? A mirror?

To sum up, the 61 yes-no questions consist of 9 checks and 52 queries, which are further subdivided into 34 utterances without a tag and 18 with a tag (9 times “o no?”, 7 times “o?”, 2 times “o qualcosa?”). With respect to the data distribution among the three proficiency levels, there is a difference between the levels B1 and B2 on the one hand and level C1 on the other hand. Of the 61 utterances in total, 21 were registered in the first level, 33 in the second and only 7 in the highest. This result shows that 22

See a.o. Martine Grice and Michelina Savino, “Map Tasks in Italian: Asking Questions about Given, Accessible and New Information”, Catalan Journal of Linguistics 2 (2003): 153-80.

184

Chapter Nine

during the test, the speakers with the highest proficiency level produced few questions and instead applied a strategy of providing explanations.23

3.2. Belgian Dutch (L1) corpus The Belgian Dutch corpus contains utterances of the same 10 learners who produced yes-no questions during the “spot-the-difference” dialogues in FL. This corpus contains read rather than semi-spontaneous speech. To have comparable data, the utterances in FL and in L1 had to be syntactically and pragmatically similar. Therefore, the L1 corpus consists of a set of utterances that were read aloud and were comparable to those recorded in FL with regard to syntactic structure and pragmatic function. The utterances were presented to the speakers on a computer screen. The presentation contained target utterances and other utterances that were used as distracters. Target utterances and distracters were presented in a random order. To counterbalance the non-spontaneity of read speech and to elicit an intonation that was as natural as possible, each utterance was accompanied by a short context that suggested to the speaker the desired interpretation. The data for L1 consist of 80 utterances produced by 10 speakers. The set of target utterances contains 8 yes-no questions with a morphosyntactic structure and a pragmatic function comparable to those of the corresponding questions found in dialogical speech in FL. The 8 yes-no questions are equally divided between queries and checks. Thus, each pragmatic type (queries versus checks) contains 4 utterances that are further subdivided in two pairs. Similar to the data for FL, the division of the queries is based on the presence of a tag. For L1, the tag is “of niet?”, which is the equivalent of the Italian tag “o no?” We chose this question tag because it is the most frequent in the FL corpus. The 4 checks are divided in two nominal phrases, which are similar to those recorded for FL, and two utterances with a finite verb. The latter are characterized by the absence of interrogative inversion of subject and finite verb. While inversion is present in queries, checks are in fact syntactically realized in the same way as declaratives. Interrogatives without inversion occur in Dutch as a means to request confirmation of a piece of information when the speaker is reasonably sure of it. Though Dutch queries are syntactically and prosodically marked as questions by the presence of inversion and of a specific tune, checks are not syntactically marked as questions. By inserting checks without inversions into our L1 corpus, we 23

Alfano, Crocco and Savy, this volume.

Yes-No Questions in Belgian Dutch (L1) and Italian (FL)

185

sought to verify whether the intonation pattern of Dutch yes-no questions would change when syntactic inversion did not take place. The target utterances for L1 have final word stress on the antepenultimate syllable or on the syllable before that one. The stressed syllable is open and has a CV or V structure. Although we selected words with phonologically voiced consonants to limit interruptions of the F0 pattern due to voiceless sounds, the regional pronunciations of some learners caused sounds to be voiceless in a number of cases. The utterances were analyzed using Praat.24

4. Results In this section, we present the results of the analysis of yes-no questions in L1 (§ 4.1.) and FL (§ 4.2.). Thereafter, we compare the productions in L1 with those in FL and show that there is evidence of prosodic transfer from the mother tongue (§ 4.3.). The analysis focuses on the global shape of yes-no questions (rising/falling contours) and on the main tonal events of the utterance. Taking as a point of reference the analysis of Standard Dutch proposed by Haan (2002) and Gussenhoven (2005), we provide a description of the yes-no question tune focusing on the nuclear pitch accent and on the boundary tone. Prenuclear accents are not taken into account.

4.1. Yes-No question tunes in Belgian Dutch (L1) Yes-no question patterns found in the Belgian Dutch corpus share clear similarities with those described in the literature on Netherlandic Dutch.25 On the basis of these similarities, we have extended the analysis proposed for Netherlandic Dutch to the Dutch that is spoken in Belgium. This choice is based on the fact that Netherlandic and Belgian Dutch, despite their differences in pronunciation, are related varieties of the same language.26 Therefore, we found it reasonable to maintain the tonal analysis proposed for Netherlandic Dutch and apply it to the same type of utterance in Belgian Dutch. It is worth noting, however, that our data suggest that there are differences in the relative frequency of the 24

Paul Boersma and David Weenink, “Praat: doing phonetics by computer” (Computer program, version 5.3.10, March 12, 2012), http://www.praat.org/. 25 Judith Haan, Speaking of questions; Carlos Gussenhoven, “Transcription of Dutch intonation”. 26 Jo Verhoeven, “Belgian Standard Dutch”, Journal of the International Phonetic Association 35 (2005): 243-47.

186

Chapter Nine

interrogative tunes in Belgian and Netherlandic Dutch. The pattern that predominantly appears in our data is reported to be marginal in Netherlandic Dutch by Haan.27 The data show a relationship between the type of yes-no question and the intonational pattern. Queries ending with a question tag and queries without a question tag are prosodically realized in different ways. Following the analysis proposed by Haan for one of the yes-no question contours in Netherlandic Dutch, we analyze the queries without a tag as L*H H%.28 We analyze the queries with a tag as L*HL L%29. All checks are produced with an L*H H% tune.30 Examples of queries without a tag are presented in Figures 9-1 and 9-2. These queries present a rising intonation contour, or a pitch rise that starts from a local F0 minimum occurring on the final word. The low turning point is analyzed as a pitch accent L* associated with the stressed syllable. The pitch rise continues until the end of the utterance, reaching a high final F0 value that almost always (in 17 cases on 19 utterances31) corresponds to the highest point of F0 for the utterance. This target point is analyzed as H%. With regard to the position of L*, in 11 cases on 19 it is aligned at the onset of the stressed vowel. Because the F0 rise starts in the initial part of the stressed syllable, we analyze the pitch accent as bitonal L*H and the complete tune as L*H H%. In a minority of cases (5), in which the pitch stays low for most of the stressed vowel and the F0 rise starts later, the pitch accent can be analyzed as L*.32

27 Haan, Speaking of questions, 111-13. Note that queries with question tag “of niet?” are not included in this study. It is thus not possible to compare the interrogative tune found in Belgian Dutch for this type of yes-no question with data for Netherlandic Dutch. 28 Ibidem. 29 Gussenhoven, “Transcription of Dutch intonation”, 135-36. 30 Haan, Speaking of questions, 111-13. 31 Of the 20 queries without tag registered for L1, 1 utterance had to be eliminated from the corpus due to narrow focus intonation. 32 Haan, Speaking of questions, 111-13.

Yes-No Questions in Belgian Dutch (L1) and Italian (FL)

187

Fig. 9-1: Heb je zin om mee te gaan naar het galabal? (Do you feel like going to the grand ball?) (proficiency level B2).

Fig. 9-2: Is dat zijn weduwe? (Is that his widow?) (proficiency level C1).

188

Chapter Nine

The queries with a question tag (Figures 9-3 and 9-4) show an intonation contour that is partly similar to that of the queries without a tag. A rising movement of F0 is also observed in this case, starting from a low pitch target (L*) aligned, in 15 cases on 18,33 at the onset of the stressed vowel and ending with a high peak (H) on the vowel of the last syllable before the tag. This peak of F0 is analyzed as a high trailing tone that forms together with L* the bitonal pitch accent L*H. On the tag, the F0 gradually falls, reaching a low level that is analyzed as a low boundary tone L%. Following the analysis of the delayed fall proposed by Gussenhoven, we hypothesize that the falling pitch movement after H corresponds to a low trailing tone L.34 Note that with the addition of the trailing tone, the pitch accent becomes tritonal (L*HL). The complete tune is then L*HL L%. In spite of their syntactic differences (simple noun phrase versus sentence with subject and finite verb), checks (Figure 9-5 and 9-6) do not display differences on the prosodic level. In the analyzed corpus, they are all produced with the same intonation contour that is observed for the queries without a tag: a rising pitch movement that starts from a local F0 minimum on the stressed syllable of the final word (L*), which is aligned, in 24 cases on 37,35 at the vowel onset of the syllable, and that ends with a high peak of F0 at the end of the utterance (H%). The nuclear pitch accent is again analyzed as bitonal L*H (L*H H%), but in a minority of cases (6) the pitch accent can be analyzed as L*. It should be noted that according to Haan, the dominant nuclear pitch accent for yes-no questions in Dutch is H*L, and L*H and L* occur only occasionally.36 The results of the analysis on Belgian Dutch do not agree because the dominant pitch accent in the analyzed corpus is L*H (45 cases on 56).37 The differences between our results and those presented by Haan could be related to intonational differences between the Dutch spoken in the Netherlands and Belgian Dutch.

33

Of the 20 queries with tag registered for L1, 2 utterances had to be eliminated from the corpus due to narrow focus intonation or interruptions in the F0 pattern of the stressed syllable. 34 Gussenhoven, “Transcription of Dutch intonation”, 135-36. 35 Of the 40 checks registered for L1, 3 utterances had to be eliminated due to narrow focus intonation or declarative intonation. 36 Haan, Speaking of questions, 111-13. 37 Note that this refers to queries without question tag “of niet?” and checks. For queries with question tag comparison is not possible. See note 30 above.

Yes-No Questions in Belgian Dutch (L1) and Italian (FL)

189

Fig. 9-3: Is dat zijn weduwe of niet? (Is that his widow or not?) (proficiency level B2).

Fig. 9-4: Heb je zin om mee te gaan naar het galabal of niet? (Do you feel like going to the grand ball or not?) (proficiency level C1).

190

Chapter Nine

Fig. 9-5: Het gaat regenen? (It is going to rain?) (proficiency level B1).

Fig. 9-6: Een adelaar? (An eagle?) (proficiency level B1).

Yes-No Questions in Belgian Dutch (L1) and Italian (FL)

191

4.2. Yes-No question tunes in Italian as FL Several features observed in Belgian Dutch question intonation are also present in the data of Italian as the FL. In the FL dialogues, there is a clear difference between queries with a tag and without a tag. An example of queries without a tag realized as L*H H% is presented in Fig. 9-7, while Figures 9-8 and 9-9 show cases of queries with a tag realized as L*HL L%.

Fig. 9-7: È un un un una donna? (Is it a woman?) (proficiency level B1).

The data on checks in Italian as the FL show that this type of yes-no question can be realized with both of the tunes observed in query moves. Neither one of the two profiles is more common in the sample (there is a balanced distribution: 4 versus 5). Examples of checks realized as L*H H% and L*HL L% are presented in Figures 9-10 and 9-11, respectively.

192

Chapter Nine

Fig. 9-8: Vedi l’ombelico o no? (Do you see the navel or not?) (proficiency level B2).

Fig. 9-9: Tu hai anche un un uomo o? (Do you also have a man or?) (proficiency level B2).

Yes-No Questions in Belgian Dutch (L1) and Italian (FL)

Fig. 9-10: La bandiera? (The flag?) (proficiency level B1).

Fig. 9- 11: Dietro? (Behind?) (Proficiency level B2).

193

194

Chapter Nine

The features of the F0 pattern of the queries without a tag (Fig. 9-7) match those of the corresponding yes-no questions in L1. The pattern presents a rising movement that starts from a local minimum of F0 on the stressed syllable (L*). The low turning point is consistently (in 29 cases on 34) aligned at the vowel onset of the stressed syllable. The pitch rise continues until the end of the utterance, and it ends with a high value of F0 corresponding to the highest pitch point in the utterance (H%). In accordance with the interpretation put forward for the same group of utterances in L1, the pitch accent is analyzed as a bitonal nuclear pitch accent L*H followed by a H% boundary tone that completes the tune. In the case of the queries with a tag (Figures 9-8 and 9-9), the same features that characterize these utterances in L1 can be observed in the corresponding group of questions in FL. The pattern is characterized by a rising movement similar to the one for the queries without a tag (starting from L*, which is, in 15 cases on 18, aligned at the vowel onset of the stressed syllable), but it does not continue until the end of the utterance. Instead, the pitch starts descending from the vowel of the last syllable before the tag, and it ends with a low F0 plateau on that tag that is analyzed as a low boundary tone L%. The high peak of F0 after L* and the falling movement towards L% is then analyzed as a high trailing tone H and a low trailing tone L. Therefore, the complete tune is L*HL L%. Checks in FL include both tunes observed in the queries.38 The two types of queries in L1 and FL display similar pitch accents (L*H versus L*HL). This observation would suggest that all queries have a basic pitch accent L*H. However, it is worth noting that according to the analysis of Gussenhoven (2005), the two accents, despite their similarity, belong to different categories: L*H is a basic pitch accent, whereas L*HL is a modified pitch accent that comes from H*L. The modification of the basic H*L consists of a delay of the peak, which creates a syllable with an L pitch accent followed by a rise-fall movement. Therefore, if the analysis of the queries with a tag as L*HL L% is correct, queries can have different pitch accents, namely L*H or H*L (delayed and prefixed = L*HL), depending on the presence of a tag.

38

Note that longer phrases (i.e. with more syllables) are produced with L*H H% (4 cases), whereas shorter ones are produced with L*HL L% (4 cases). In one case the longer phrase is also produced with L*HL L%.

Yes-No Questions in Belgian Dutch (L1) and Italian (FL)

195

4.3. Prosodic transfer from L1 to FL The analysis of the data for L1 and FL shows that the intonation contours produced in FL share several features with those produced in L1. In their mother tongue, the speakers produced yes-no questions using the tune L*H H% for queries without a tag and checks comprised of a simple noun phrase or a sentence without syntactic inversion. The speakers also produce an L*HL L% tune for queries followed by a tag. These two tunes have also been found in the productions in FL, although in the productions of checks, their distribution does not always correspond to the distribution observed in the mother tongue. In L1, checks are realized as L*H H%, but in FL, both L*H H% and L*HL L% occur. Therefore, the data suggest that the learners apply the same intonational choices in FL as they do in L1. The fact that various similarities are observed between L1 and FL indicates that the learners transfer their native intonation to FL. Given the similarities between the tunes observed in L1 and FL, the transfer seems to primarily involve the selection of the phonological categories (nuclear pitch accent and the boundary tone). Furthermore, the visual analysis of the 61 utterances in FL proposed in this paper suggests that the speaker’s performance does not change appreciably in the three groups of learners. However, the measurements presented by Alfano, Crocco and Savy in this volume show that the pitch rise starts significantly later in the group of learners with a higher competence level (C1) than in the two groups of beginners (B1 and B2).39 This result shows that the advanced learners have a tendency to approximate the intonation of yes-no questions produced by native speakers of Italian. In the varieties of Italian that form the input provided to the students by their lecturers, the question rise begins after the stressed syllable; therefore, a later rise in the FL can be interpreted as an approximation of the natives’ intonation.40 Based on the observations presented in this paper and on the analysis performed by Alfano, Crocco and Savy, we can conclude that new learners transfer their native interrogative intonation to the FL, whereas in the interlanguage of the advanced learners, the prosodic patterns tend to be modified to realize a better approximation of the interrogative intonation of native speakers.

39 40

Alfano, Crocco and Savy, this volume. Ibidem.

196

Chapter Nine

5. Conclusions In this paper, we provided a description of the yes-no question intonation in Belgian Dutch (L1) and a comparison between the question intonation in L1 and the corresponding intonation in Italian as FL. We applied the intonational analysis proposed by Haan (2002) and Gussenhoven (2005) for Netherlandic Dutch to the interrogative contours of query and check in Belgian Dutch to show that in the Belgian Dutch corpus, the most frequent yes-no question tune is L*H H%. This question tune was marginal in the corpus of Netherlandic Dutch examined by Haan.41 The analysis of Belgian Dutch also shows that differences on the syntactic level (i.e. presence versus absence of a question tag) correspond to differences at the intonational level. Queries without a question tag are realized as L*H H%, whereas the corresponding utterances with a tag present another tune, L*HL L%. However, no clear difference has been found between utterances with and without inversion of the subject and verb. The comparison between the productions in Belgian Dutch (L1) and those in Italian (FL) provided preliminary evidence of prosodic transfer. The transfer primarily involved the selection of the tonal categories (pitch accents L*H and L*HL, and boundary tones L% and H%). From the phonetic point of view, there is evidence that the low turning point (L) at which the interrogative rise begins occurs later in the group of advanced learners (proficiency level C1) than in the group of beginners (proficiency levels B1 and B2).42 This finding suggests that advanced learners modify their productions as they try to produce a more native-sounding question intonation.

41 42

Haan, Speaking of questions, 111-13. Alfano, Crocco and Savy, this volume.

PART III PROSODY AND PRAGMATICS

CHAPTER TEN RYTHM IN NARRATION AND DESCRIPTION IN L2 ITALIAN MARTA MAFFIA, ANNA DE MEO AND MASSIMO PETTORINO UNIVERSITY OF NAPLES “L’ORIENTALE”, ITALY

1. Introduction Since the last decades of the nineteenth century, task-based language teaching, i.e. the use of tasks aiming at the realization of linguistic acts to achieve a specific communicative outcome, has been widely used in the teaching of a second language1. This approach allows the learner to focus on the communicative functions of the language rather than on the accuracy of the linguistic forms. In first language acquisition, the skill of producing efficient communicative language acts normally relies on the innate capacity of the adult speaker, whereas the learning of a second language requires the acquisition of norms regulating the pragmatics of the new language besides mastering other linguistic devices (morphosyntactic, lexical, etc) needed for an adequate encoding of the linguistic acts.

1

Jane Willis, A framework for task-based learning (Harlow: U. K. Longman Addison-Wesley, 1996); Rod Ellis, Implicit and explicit learning of language (London: Academic Press, 1994); Rod Ellis, Task-based Language Learning and Teaching (Oxford: Oxford University Press, 2003); Martin Bygate, “Quality of language and purpose task: pattern of learners’ language on two oral communication tasks”, Language Teaching Reasearch 3/3 (1999): 185-241; Tracey M. Derwing et al., “Second Language Fluency: Judgements on Different Tasks”, Language Learning 54 (2004): 655-79.

200

Chapter Ten

As many studies have shown, the type and the complexity of the linguistic task can undoubtedly affect the communicative outcome in L2.2 In this study, we investigate some prosodic aspects of the narrative speech and of the descriptive speech. By their very nature, narrations and descriptions differ both in linguistic forms and in their use of cognitive and discourse management instruments.3 In a narrative task, time is the conceptual domain at the basis of discourse organisation. The speaker tends to employ as guiding scheme the temporal succession of events, unless he or she wishes to produce particular effects on the interlocutor. This discourse structure allows the speaker to produce completely comprehensible narration texts in L2, efficient from the communicative point of view, even with poor linguistic means.4 The temporal scheme, in a way compulsory for the development of the events, is, however, missing in the description task. It is, in fact, more difficult to identify a unique domain guiding the organization of the descriptive text, and the speaker is forced to choose among various possible organizational discourse models, based on diverse parameters. Besides, the possibility of structuring a description and the modality used inevitably depend on the characteristics of the referent in mind: it is one 2

Many tasks and linguistics acts, of more or less complexity, have been analyzed in the literature in different languages: narrations, descriptions, refusals, compliments, excuses, complains, etc. 3 Christiane von Stutterheim, “Global principles of information organisation in texts of L2 speakers”, in Studi Italiani di Linguistica Teorica e Applicata, Strutture testuali e principi di organizzazione dell’informazione nell’apprendimento linguistico, eds. Marina Chini and Anna Giacalone Ramat (Roma: Pacini Editore, 1998), 89-110; Wolfgang Klein and Christian v. Stutterheim, “Questio und referentielle Bewegung in Erzählungen”, Linguistische Berichte 109 (1987): 163-83. 4 In a narration, moreover, direct speech can be a useful mean of presenting events with efficacy, avoiding the use of complex indirect syntactic structures. Cf. Marina Chini, “Testualità e mezzi referenziali concernenti la persona in narrazioni di italofoni e di apprendenti di italiano L2”, in Studi Italiani di Linguistica Teorica e Applicata, Strutture testuali e principi di organizzazione dell’informazione nell’apprendimento linguistico, eds. Marina Chini and Anna Giacalone Ramat (Roma: Pacini Editore, 1998), 153-81; Marina Chini, ed. Topic, struttura dell’informazione e acquisizione linguistica (Milano: Franco Angeli, 2010); Ruth Berman and Dan I. Slobin, Relating events in Narrative: a crosslinguistic developmental study (Hillsdale: Lawrence Erlbaum Associates Publishers, 1994); Maria G. Lo Duca, “Sulla rilevanza per la glottodidattica dei dati dell’acquisizione di lingue seconde: ‘narrare’ in italiano L2”, in Verso l’Italiano, ed. Anna Giacalone Ramat (Roma: Carocci, 2003), 254-70.

Rhythm in Narration and Description in L2 Italian

201

thing to describe an action or a process, such as the functioning of an instrument, and quite another is to describe a fixed, still object. This task is thus generally more complex from the cognitive point of view and might be limiting the exposition, especially when the basic competence in L2 does not allow an adequate level of information explicitness and precision.5 In the Common European Framework of Reference for Languages (CEFR) both description and narration are considered skills to be developed from the early stages of language learning, although limited to specific areas of use.6 According to the CEFR, the learner should already be able to describe him- or herself and some aspects of his or her life at the A1 level of language competence. At the A2 level, the learner possesses “a simple descriptive language” which enables him or her to describe aspects of everyday life in familiar situations and to describe and compare two objects. Only at more advanced levels learners are able to describe fluently and in details a greater range of elements, also complex, taking distance from him- or herself and from the surrounding world (Table 10-1). In narrations, on the other hand, the area of use seems to expand earlier and easier (Table 10-1). The ability to narrate occurs at A2, the elementary level, (“Can you tell a story or describe something in a simple list of points?”), but at level B1 the ability of narrating a story, the plot of a novel or of a film is already expected (“can narrate a story”, “can relate a plot of a book or film”.). Therefore, while in narrations it seems to be easier to go beyond relating on personal experience and every day life, in descriptions the learner needs to reach level C2 of language proficiency in order to be able to produce “clear, smoothly flowing, elaborate and often memorable descriptions”. The aim of our study is to analyse and compare the prosodic features of descriptions and narrations in L1 and L2 Italian. Our initial question was whether and in what way the diverse nature of the two linguistic tasks might influence the prosodic parameters of these two productions. Moreover, we wished to look into whether in the performance of these 5

See also Mary Carroll, “Spatial information in descriptions: how advanced L2 speakers (L1 English- L2 German) manage coherence”, in Studi Italiani di Linguistica Teorica e Applicata, Strutture testuali e principi di organizzazione dell’informazione nell’apprendimento linguistico, eds. Marina Chini and Anna Giacalone Ramat (Roma: Pacini Editore, 1998), 183-200. 6 Council of Europe, The Common European framework of reference for languages: Learning, teaching, assessment (Cambridge: Cambridge University Press, 2001).

202

Chapter Ten

type of tasks in L2 Italian, at different levels of L2 linguistic competence, and in different mother tongues, we could find linguistic behaviours analogous to those present in the L1 performance. SUSTAINED MONOLOGUE: Describing Experience Can give clear, smoothly flowing, elaborate and often memorable C2 descriptions. Can give clear, detailed descriptions of complex subjects. C1 Can give elaborate descriptions and narratives, integrating sub themes, developing particular points and rounding off with an appropriate conclusion. Can give clear, detailed descriptions on a wide range of subjects B2 related to his field of interest. Can give straightforward descriptions on a variety of familiar B1 subjects within his field of interest. Can reasonably fluently relate a straightforward narrative or description as a linear sequence of points. Can give detailed accounts of experiences, describing feelings and reactions. Can relate details of unpredictable occurrences, e.g. an accident. Can relate the plot of a book or film and describe his/her reactions. Can describe dreams, hopes and ambitions. Can describe events, real or imagined. Can narrate a story. Can tell a story or describe something in a simple list of points. A2 Can describe everyday aspects of his environment e.g. people, places, a job or study experience. Can give short, basic descriptions of events and activities. Can describe plans and arrangements, habits and routines, past activities and personal experiences. Can use simple descriptive language to make brief statements about and compare objects and possessions. Can explain what he/she likes or dislikes about something. Can describe his/her family, living conditions, educational background, present or most recent job. Can describe people, places and possessions in simple terms. Can describe him/herself, what he/she does and where he/she A1 lives. Table 10-1: Analytic descriptors of spoken production of the Common European Framework of Reference for Languages.

Rhythm in Narration and Description in L2 Italian

203

2. Materials and Methods In order to achieve our aims, a corpus of semi-spontaneous speech in Italian L1 and L2 was created. We selected four speakers, all male, residing in Naples, south Italy, aged between 27-40. Two were mother tongue Italians (N1 and N2), and the other two participants were learners of L2 Italian with different L1 and different level of linguistic competence: NN1 - L1:Wolof; level: B1, NN2 – L1: Swahili; level: A2. An assignment consisting of three different linguistic tasks was given to each of the four selected participants. The first task (task 1) asked for the description of a colour photo of a seagull in front of a closed door. The choice of this image was dictated by the need to provide as neutral an image as possible, unlikely to create any particular emotion in the observer, and illustrating easily identifiable objects with well defined colours and clear space distribution. The second task (task 2), a visually guided storytelling, required the production of a short narration, elicited by a comic strip. This story was chosen becauseit was short and simple: there was no dialogue, and only two characters were present: a dog and a bird. This task can be defined as a mixed-ability task as it stands half way between the description of a series of images and the recounting of a story. The third task (task 3), narrating a personal experience, had the purpose of stimulating the speakers’ emotional involvement. The two native Italians, young, with no permanent jobs, were asked to narrate a particularly unpleasant personal anecdote experienced while looking for work. The two learners of L2 Italian, both illegal immigrants, were invited to recount their journey to Naples. In choosing the topics for the narration we were aware that both themes might possibly trigger negative emotions, like sadness, anger, or fear. Therefore, before proposing the “emotional” task, the participants were reassured that they were free to perform the task as they liked, even to the point of remaining silent or refusing the task altogether. In order to provide a natural conversational environment the recordings were made in a silent (but not soundproof) room. The 12 productions obtained were transcribed and then analysed using open source

Chapter Ten

204

Wavesurfer 1.8.8, which allowed us to estimate the speech chain length, and the number of silent and non-silent pauses actually produced.7 From these data the following rhythmic-prosodic indexes were obtained: -

Articulation Rate AR (syll/s) Speech Rate SR (syll/s) Frequency of Pauses FP (syll) Tonal Range TR (st)

A total of 44 min. speech was analysed and 7618 syllables computed.

3. Data analysis Comparing the AR and the SR mean values obtained in the three tasks (Fig. 10-1), it can be seen that while the AR values are rather stable, the SR tends to grow in task 3. This implies a different use of silent pauses either in frequency or in duration. As regards the occurrence of silent pauses, it can be observed that they become less frequent in task 3. In spoken L1 we notice a silence every 7.5/8 syllables in tasks 1 and 2, while in task 3 a silence occurs every 9.5 syllables. By contrast, in L2 speech we find a pause every 5.6 syllables in tasks 1 and 2, whereas in task 3 a pause appears every 7.5 syllables. As for the duration of silent pauses, we observe that they are more stable in L1 than in L2 (Fig. 10-2). However, in both cases, the highest value regarding task 1. Moreover, all the speakers show an increase in the percentage of articulated sequences in the two narration tasks as opposed to the description task. Therefore in task 3, the personal narrative, the speakers being more emotionally involved, tend to reduce the moments of silence, both in number and in duration. This datum also points out that in personal narration, the speaker finds less difficulty in planning his or her speech, evidently due to the nature of the task, which allows greater freedom in textual weaving.

7 Speech chain is considered that portion of an utterance between two silent pauses. See Emanuela Magno Caldognetto, Enrica De Zordi and Loredana Corrà “Il ruolo delle pause nella produzione della parola”, in Il Valsalva- Bollettino della Società Italiana di Audiologia e Foniatria, 5 (1982): 12-21. In this study we consider “silent pause”, a silence interval of 0.1 s.

Rhythm in Narration and Description in L2 Italian

205

Fig. 10-1: Articulation rate and speech rate mean values in the three tasks in L1 and L2 Italian.

Fig. 10-2: Mean duration of silent pauses in the three tasks in L1 and L2 Italian.

Furthermore, as regards the use of non-silent pauses—i.e. vocalizations, lengthening, false starts, etc.—a difference between L1 and L2 can be noted. Stable values are found in the native speakers (c. 13% of the locution time), and more task-dependant values in the non-native speakers,

206

Chapter Ten

from 8% of task 1 to 13% of the utterance time in the tasks 2 and 3. As for the type of non-silent pauses, lengthening is mostly used by all speakers (from 60% to 70% of the total). Furthermore, it is interesting to note that in task 2, the range of non-silent pauses are reduced with respect to the other two tasks. In the description task, therefore, the foreign speakers display a low percentage rate in articulation time (32%), many silences (60%) and few disfluencies. It is likely that the rigidity of the task matched with a limited linguistic competence, resulting in a tendency to silence. On the other hand, in the two narratives, as the amount of possible content to express grows and, in task 3, it increases along with the greater freedom in textual organization, an increase of the percentage rate in speech articulation together with more disfluencies are detected, as a consequence of the speaker’s desire to communicate. As far as intonation movement is concerned, the data obtained do not seem to highlight a univocal relationship between tonal range and the type of task in either L1 or L2. As regards the impact of the different levels of competence on the speech quality of the two non-native participants, our data show that at the lowest level (A2) the tendencies registered in the native speakers’ speech are emphasized. It is interesting to highlight the diverse elocution time composition in the three different tasks (Fig. 10-3). The speaker of A2 level (NN2), due to his difficulties in organizing his discourse with adequate vocabulary, shows in the first two tasks a silence percentage between 59% (task 2) and 68% (task 1) in total utterance time, while in task 3, in which he feels more free and involved, his silence decreases to 33%. Language difficulties are also evident in the analysis of the mean duration of silent pauses, which for this speaker are directly linked to the type of task: just over 1 s. for the less bound task 3, the narration, while over 3 s. for task 1, the description. These data show that the level of linguistic competence in a second language is not exhausted in the acquisition of a skill, but that it corresponds also to a lower or higher degree of language production stability. It is reasonable to assume, therefore, that a higher level of competence guarantees more stable values, regardless of the type of task.

Rhythm in Narration and Description in L2 Italian

Articulated sequences

Silent pauses

207

Non-silent pauses

Fig. 10-3: Utterance composition: articulated sequences, silent pauses and nonsilent pauses.

208

Chapter Ten

4. Communicative task and emotions perceived Given the particular nature of the third task, which presumes, more than the other two, a high level of emotional involvement, we were interested in finding out whether some kind of emotion would indeed be perceived in this task as compared with the other two. By “emotion” we mean that complex subjective condition involving the physiological, kinaesthetic, motivational aspects of the self.8 In this work we have focused our attention on the rhythmic-prosodic aspects of emotional speech and left out other aspects affecting it, such as the lexical, the syntactic, and the morphological ones. In order to evaluate emotion perception we devised a perception test using audio-fragments of 6-7 s. from the 3 tasks performed by the 4 interlocutors (6 fragments per speaker). The audio files were randomly assembled into one listening file which was then administered to 49 native listeners: 29 women and 20 men, average age 35, many of them members of a choir and thus used to listening to and identifying sounds. The listeners were asked to indicate whether each audio segment was emotionally characterized, and if so, to identify freely the type of emotion perceived (no prior indication as to the type of emotion to be recognized was provided). Moreover, in order to have listeners evaluate the emotion perceptions exclusively through prosodic parameters, the audio segments had been masked using a lowpass filter, which allows the listener to perceive only the rhythm of the utterance and not to identify its textual content. The test has confirmed that emotions are perceived even in very short speech fragments regardless of their semantic content. Although not all of the fragments had been identifiedm, as expected, there was a certain uniformity in the answers given by the listeners. To identify the emotion prosodic correlates, two segments, recognized as expressing emotions by a the majority of the listeners (NN1: 92% and N1: 76%) were selected, and their rhythmic-prosodic characteristics compared with those of the two fragments from the same interlocutors which had been previously clearly judged as lacking emotion. 8

Isabella Poggi and Emanuela Magno Caldognetto, “Il Parlato emotivo. Aspetti cognitivi, linguistici e fonetici”, in Atti del Convegno “Il parlato emotivo” Napoli, 12-15 febbraio 2003 (Napoli: D’Auria Cd Rom, 2003); Emanuela Magno Caldognetto, Federica Cavicchio and Piero Cosi, eds., Comunicazione parlata e manifestazione delle emozioni. Atti del 1° Convegno Nazionale GSCP – Gruppo di Studio della Comunicazione Parlata, Padova 30 Novembre - 1 Dicembre 2004 (Napoli: Liguori, 2008).

Rhythm in Narration and Description in L2 Italian

209

It is possible to notice (Table 10-2) that the two audio-fragments labelled as “emotional” present, in fact, different characteristics: with respect to the neutral fragment, NN1 shows lower values all through except in the tonal range which remains more or less stable; whereas N1, shows higher values with silent pauses becoming shorter, on average.

AR SR FP Mean Pause Duration TR

NN1 neutral 6.51 4.86 11

NN1 emotional 4.44 1.99 6

N1 neutral 5.8 4.7 8.5

N1 emotional 6.79 6.29 20.5

1.03

0.8

0.41

0.1

7.70

8.42

5.37

12.16

Table 10-2: Rhythmic-prosodic characteristics of audio fragments labeled as “emotional” and “neutral”. If we compare the above data with the parameters found in the literature for primary emotions,9 it is possible to verify that, except for the tonal range, the emotional fragments present rhythmic-prosodic characteristics which are associated with particular emotional models. The first fragment presents the characteristics of low activation emotion, whereas the second of high activation emotion.

9

Klaus R. Scherer, “Vocal communication of emotion: A review of research paradigms”, Speech Communication 40 (2003): 227-56; Marc Schröder, “Speech and Emotion Research: An overview of research frameworks and a dimensional approach to emotional speech synthesis” (PhD thesis, Saarland University, 2008); Massimo Pettorino, “Inglese, italiano e giapponese: analisi dei correlate acustici delle emozioni nel parlato cinematografico”, in Comunicazione parlata e manifestazione delle emozioni. Atti del 1° Convegno Nazionale GSCP – Gruppo di Studio della Comunicazione Parlata, Padova 30 Novembre - 1 Dicembre 2004, eds. Emanuela Magno Caldognetto, Federica Cavicchio and Piero Cosi (Napoli: Liguori, 2008), 45-57; Ilaria Vitagliano, Antonella Giannini and Massimo Pettorino, “Le emozioni nel tempo”, in Fenomeni di intensità nell’italiano parlato, eds. Barbara Gili Fivela and Carla Bazzanella (Firenze: Franco Cesati, 2008); Vincenzo Galatà and Luciano Romito, “Un corpus sperimentale per lo studio cross-linguistico europeo delle emozioni vocali”, in Proceedings of the 5th AISV National Conference - La dimensione Temporale del parlato, Università di Zurigo, Svizzera, 4-6 February 2009, eds. Stephan Schmid, Michael Schwarzenbach and Dieter Studer (Torriana, RN: EDK, 2010), 603-42.

210

Chapter Ten

Feedback from the test confirms the above data. In fact, the first fragment was labelled as “sad” by the majority of the listeners (Fig. 10-4), and the second as “anxious” (Fig. 10-5).10 Considering that the listeners were completely free in giving their evaluations and consequently their choice of labelling the emotions was potentially unlimited, the results obtained are significant.

Fig. 10-4: Emotions perceived in the speech of NN1.

Fig. 10-5: Emotions perceived in the speech of NI. 10

In psychology anxiety is associated with fear, high activation emotion.

Rhythm in Narration and Description in L2 Italian

211

It is interesting to note that the audio-fragment of NN1, recognized as emotional, is in fact part of the descriptive text (task1), while that produced by N1 is instead part of the task 3. These data demonstrate that there is no absolute correspondence between the type of task and the presence or absence of emotions, at least from the perceptive point of view. Moreover, the hypothesis according to which “sadness and anger are generally best recognized vocally, followed by fear” seems to be confirmed.11 According to this theory, during the evolution of the human being, some expressions of emotion (among which, anger, sadness, and fear) are supposed to have developed almost exclusively vocally, while other emotions, such as joy and disgust, would have been transmitted through other instruments such as facial expressions.12

5. Conclusions The analysis of the prosodic parameters characterizing spontaneous speech in description and narration in the three specific tasks in L1 and L2 Italian (describing a photograph, visually guided storytelling, personal narrative) confirms the findings from previous studies13 on articulation rate, speech rate and fluency, in free production performance and image description in L1 Italian. As regards L2 Italian, however, our study has highlighted a tendency for longer and more frequent silences in performing descriptions, especially at the lower levels of linguistic competence. This tendency can be attributed to the greater difficulty that this type of task poses on speakers in the management of discourse planning, information organization, and lexical choice. By contrast, in personal narratives, in which the speakers had been emotionally involved the most (task 3), a tendency towards increased fluency has been registered. The wish to narrate and of being personally involved in a story along with a greater 11

Klaus R. Scherer, Tom Johnstone and Gudrun Klasmeyer, “Vocal expression of emotion”, in Handbook of the Affective Sciences, eds. Richard J. Davidson, Klaus R. Scherer, Hill Goldsmith (New York and Oxford: Oxford University Press, 2003), 444. 12 According to this study, in the first case the need of communicating threat and danger, even from great distance, had prevailed; in the second case, that of immediate understanding was predominant. 13 See e.g. Emanuela Magno Caldognetto and Kiki Vagges, “Indici di fluenza, tipologia e distribuzione delle sillabe nel parlato spontaneo”, in Atti del XIX Convegno nazionale AIA (Napoli: 1991), 423-29.

212

Chapter Ten

freedom allowed in structuring the discourse and selecting the topics, seem to encourage longer speech chain use both in L1 and L2. In the productions of the two non-native speakers this is particularly evident in the speaker with a lower level of linguistic competence. As regards nonsilent pauses, it is worth noticing a great stability, both qualitative and quantitative, found in L1, in contrast with the non-native speakers’ narrations. In particular, for the L2 speakers, the number of non-silent pauses tends to increase in task 3. Furthermore, our study data have highlighted a strong relationship between the degree of rhythmic-prosodic variability in the utterance and the level of linguistic competence in L2: a lower level of competence corresponds to greater diversification in rhythm and prosody, relatively to the degree of difficulty the interlocutor finds in performing the different tasks. The results of the perception test, based on short speech sequences masked with a lowpass filter, have confirmed that prosodic traits can be characterized as a “code, having autonomous communicative efficiency with respect to the verbal code”, a code, therefore, capable also of conveying emotions.14 The emotions recognized by the listeners in our study correspond to the models for primary emotions found in the literature. In the two audio-fragments perceived as “emotional”, the listeners had identified two diverse types of emotion: one of low activation (sadness) and one of high activation (anxiety). The prosodic parameters of these fragments correspond to the patterns for the elocution rhythm reported the literature: i.e. AR, SR and silent pause frequency higher for high activation emotions and lower for low activation emotions. But as regards the intonation movement, data from the present study show a direct general relationship between the tonal range spread and the degree of emotional feeling perceived, independently of the type of evaluation given. Finally, the perception test, has ascertained that it is not always justified to consider an utterance as globally emotional or neutral, as emotions can affect even very limited parts of the utterance, and, in the display of different parameters, assume the variety of colouring characterizing spontaneous speech.

14 Luigi M. Anolli and Rita Ciceri, La voce delle emozioni. Verso una semiosi della comunicazione vocale non-verbale delle emozioni (Milano: Angeli, 1992), 387.

CHAPTER ELEVEN FOREIGN ACCENT AND PERSUASIVENESS: NATIVE AND NON-NATIVE VOICES IN A RADIO SPOT MARILISA VITALE, ANNA DE MEO AND MASSIMO PETTORINO UNIVERSITY OF NAPLES “L’ORIENTALE”, ITALY

1. Introduction In radio advertising, the way the message is spoken is the only route the listener, thus voice, beside textual content, is key to effective communication. Different voices can lead to different attitudes in listenersconsumers, since the specific characteristics of the spokesperson’s voice and speech could determine its persuasiveness.1 Several studies carried out in different fields of research (marketing studies, social psychology, linguistics) have shown that non-standard and foreign accented speech are 1

Oscar W. DeShields, Jr., Ali Kara and Erdener Kaynak, “Source effects in purchase decisions: The impact of physical attractiveness and accent of salesperson”, International Journal of Research in Marketing 13 (1996): 89-101; Susannah V. Levi and David B. Pisoni, “Indexical and linguistic channels in speech perception: Some effects of voiceovers on advertising outcomes”, in Psycholinguistic Phenomena in Marketing Communications, ed. Tina M. Lowrey (Mahwah: Lawrence Erlbaum Associates, 2007), 203-19; Rachel Maldonado, Patriya Tanshuaj and Darrel D. Muehling, “The impact of gender on Ad processing: A social identity perspective”, Academy of Marketing Science Review 3 (2003): 1-15; Richard E. Petty and John T. Cacioppo, “Issue Involvement as a Moderator of the Effects on Attitude of Advertising on Content and Context”, Advances in Consumer Research 8 (1981): 20–4; Thomas W. Whipple and Mary K. McManamon, “Implications of using male and female voices in commercials: An exploratory study”, Journal of Advertising 31 (2002): 79-91.

214

Chapter Eleven

often unfavourably evaluated, since they convey important clues about the speaker’s social group, whose stereotyping may negatively influence the way voice is perceived.2 However, the association of the marketed product with the speaker’s social group is sometimes fruitfully exploited in order to create a positive mental association. When dealing with goods or services strongly connoting a particular culture or a specific gender group, the use of an accented voice may be a useful strategy to attract the consumer. As a matter of fact, voice-related prejudices and attitudes are usually projected on the entire advertisement, causing the success or failure of the intended goal. In addition, it should be noted that familiar voices are preferred to “strange” ones, since they activate a kind of interpersonal attraction due to the idea that people sharing the same cultural and social background tend to hold the same point of view.3 Although many studies have been carried out on variables influencing persuasive communication,4 very few focus on listeners’ reaction to advertisements using L2 speech. For this reason, the present study aims to investigate the relationship between a foreign accent and persuasiveness in the Italian language. To this end, a perceptual test was set up, involving both native and non-native Italian speakers, in order to evaluate native

2

Dawn Birch and Janelle McPhail, “The impact of accented speech in International television advertisements”, Global Business Languages 2/9 (2010): 91-105; Oscar W. DeShields, Jr. and Gilberto de los Santos, “Salesperson’s accent as a globalization issue”, Thunderbird International Business Review 42 (2000): 29-46; Caroline Jones, Lynn Berry and Catherine Stevens, “Synthesized speech intelligibility and persuasion: Speech rate and non-native listeners”, Computer Speech and Language 21 (2007): 641-51; John Tsalikis, Oscar W. DeShields, Jr. and Michael S. LaTour, “The role of accent on the credibility and effectiveness of the salesperson”, Journal of Personal Selling and Sales Management 11 (1991): 31-41. 3 Carl I. Hovland, Irving L. Janis and Harold H. Kelley, Communication and Persuasion (New Haven: Yale UP, 1953). 4 Judit A. Hall, “Voice and tone persuasion”, Journal of Personality and Social Psychology 38/6 (1980): 924-34; David E. Kanouse and Robert P. Abelson, “Language variables affecting the persuasiveness of simple communication”, Journal of Personality and Social Psychology 7/2-1 (1967): 158-63; Stephen M. Smith and David R. Shaffer, “Celerity and Cajolery: Rapid Speech May Promote or Inhibit Persuasion through its Impact on Message Elaboration”, Personality and Social Psychology Bulletin 17 (1991): 663-69; Stephen M. Smith and David R. Shaffer, “Speed of Speech and Persuasion: Evidence for Multiple Effects”, Personality and Social Psychology Bulletin 21 (1996): 1051-60.

Foreign Accent and Persuasiveness

215

listeners’ reactions to foreign accented speech, and to correlate them with the spectro-acoustic analysis values.

2. Perceptual test For the purpose of this study two native Italian speakers and six speakers of L2 Italian, all non-professional spokespeople, were selected. As regards the former, in order to choose standard Italian speakers without any regional inflection, a pre-test was arranged and administered to 70 native Italian listeners: a short read text was recorded by 10 speakers, both male and female, aged between 20-30, and a listening tape with the files mixed in random order was prepared. For each item, informants were asked to: 1) indicate if the speaker was native or not; and 2) evaluate the degree of regional accent on a three-point-scale (1= standard accent / 3= strong regional accent). To prevent results from possibly being affected by diatopic factors, both speakers and listeners had been selected from the Campania region. Two speakers recognized as native Italians by approximately 100% of the listeners and also judged as having a standard Italian accent (88%) were finally chosen for the experiment. As for the six non-native speakers, they had been chosen on the basis of their mother tongue and of their competence level of the Italian language. Specifically, to get speech productions devoid of all disfluencies, errors or other features which might have negatively affected comprehensibility and consequently persuasiveness, all the involved foreign speakers had been chosen as having a C1 level of competence of L2 Italian (CEFR). Regarding their mother tongue, 2 French, 2 Russian and 2 Chinese speakers had been involved in the study on the basis of the different degree of familiarity Italians have with these foreign languages, for historical and geographical reasons. To avoid gender-dependent results, for each considered L1, both male and female speakers had been chosen. The corpus, consisting of a radio advertisement, was constructed so to avoid being affected by social identification phenomena in its processing and evaluation. As for the theme, a social campaign designed to increase awareness of multiple sclerosis was used. This kind of advertisement had been chosen on basis of its total disconnection with every kind of genderconnoted product. In traditional radio or television commercials, the use of a male or a female voice could, in fact, affect the degree of persuasiveness, in relation to the type of marketed product. On the contrary, as shown by some studies conducted in Anglo-Saxon circles on the voice-genderadvertisement relationship, social advertisements, having a non-commercial

216

Chapter Eleven

goal, and thus being far from standard marketing practices, are devoted to the whole social community and not to a specific gender group. The spot is divided into two parts: it opens with information and ends with a persuasive message. For our purpose, the spot, originally intended for television broadcasting, had been deprived of the visual component (consisting of a clock with hands moving) and was transformed into a radio commercial in order to isolate the “voice” element and check its specific influence on the perception of the message. The 8 recorded radio spots were then assembled into one listening file administered to 164 native Italian speakers, 45 males and 116 females, with a mean age of 17 years, all in the last two years of high school. To prevent results from being influenced by diatopic variables, the listeners were also from the Campania region. Each listener was asked to: 1) assess the degree of foreign accent by tagging each voice as having a “strong foreign accent”, “mild foreign accent” or “native accent”; 2) evaluate the degree of persuasiveness of each voice on a three point scale (1=poor, 2=sufficient, 3=good); and 3) indicate one or more parameters that most influenced their judgment, whether positive or negative, choosing among articulatory accuracy, intonation, speech rate, pauses or other parameters. The administration session had been preceded by an explanation of the evaluation form to be used and its filling criteria, especially regarding the aforesaid parameters. The test was proposed as a study aiming at selecting the best voice for a radio spot, so that listeners would focus as much as possible on the persuasiveness of the message rather than on the foreign accent of the speaker.

3. Perceptual test results 3.1. Foreign accent As for the foreign accent assessment, the results of the perceptual test show that only the two Italian spokespeople were recognized as native speakers by almost all of the listeners. On the contrary, despite the high level of Italian language competence, 5 of the 6 non-native speakers were judged as foreign accented by nearly 100% of the listeners, while only the male Russian speaker was assigned a small percentage of “native” ratings (24%). Moreover, differently from all other non-natives, prevalently judged as having a “strong foreign accent”, the French spokeswoman and the Russian spokesman were evaluated as “mild” foreign accented (Fig. 11-1).

Foreign Accent and Persuasiveness

217

Fig. 11-1: Degree of foreign accent: percentage values.

3.2. Persuasiveness Since advertisements have a well-defined connative function, i.e. that of inducing the addressee to do or think what the speaker wishes, we decided to focus our analysis only on the judgment of “good” persuasiveness, omitting the “sufficient” one. In this perspective, the general data concerning persuasiveness clearly indicate that for Italian listeners the most effective speakers are the native ones. This result seems to prove a direct relationship between the degree of perceived foreign accent and the judgment of persuasiveness (Fig. 11-2).

Fig. 11-2: Degree of persuasiveness. L1 and L2 Italian mean values (%).

218

Chapter Eleven

However, since the French woman was the only foreigner to have obtained a positive evaluation, deviating from the overall negative nonnative values (non-native range = 0-3%; French woman = 42%), we decided to investigate in detail the acoustic features characterizing the corpus, in order to verify whether there were acoustic parameters that might have determined persuasiveness assessment (Fig. 11-3).

Fig. 11-3: Percentage of persuasiveness: individual values.

Persuasiveness data were also examined from a gender-related perspective to see whether male and female listeners’ responses varied according to the gender of the spokesperson. The analysis, however, did not reveal significant outcomes, since data always appeared uniform and consistent among male and female gender groups.

3.3. Parameters Regarding the parameters that guided the judgment on persuasiveness, data analysis shows that the elements that had the greatest impact on the listeners’ perception were intonation and articulatory accuracy, while speech rate and pauses were judged secondary. The chart in Fig. 11-4, which shows the mean percentage values of each considered parameter, points out a perfect correspondence of the L1 and L2 assessment, thus allowing the inference that ratings were not random and that, whether positive or negative, the evaluation had not been accent-dependent.

Foreign Accent and Persuasiveness

219

As a consequence of the cross-analysis of foreign accent and persuasiveness data, which suggested the presence of some anomalies in the French female speaker’s production, it seemed useful to break up data concerning her production (Fig. 11-4). The aim was to highlight all the possible differences and/or similarities with the rest of the L2 group that may have positively orientated the listener’s evaluation on persuasiveness. In fact, the traits that have most affected listeners’ assessment for this speaker are partially different from those considered for the other nonnative spokespeople. In this case, what had most struck the native listeners was intonation, with an impact greater than that shown in all other cases, and the breaks management. We can infer, then, that pauses and pitch contour movements, if managed in the appropriate way, promote a good level of persuasiveness, despite the presence of a foreign accent.

Fig. 11-4: Degree of parameters relevance for the assessment of persuasiveness.

In summary, although there seems to be a direct relationship between the degree of perceived foreign accent and persuasiveness, yet there are also some clues suggesting further and more technical investigation into the role played by each rhythmic-prosodic feature in the evaluation of a persuasive utterance. For this reason, a spectro-acoustic analysis was carried out to identify the specific features that allowed Italian native speakers and the French female spokesperson to be more effective than the other speakers.

Chapter Eleven

220

4. Spectroacoustic analysis

Male

Female

Speech analysis was conducted on rhythmic-prosodic parameters using WaveSurfer 1.8.8 software. Speech chains duration, the number of syllables per speech chain, silent pauses duration, and maximum and minimum F0 per speech chain were measured. On the basis of the values obtained, articulation rate AR (syll/s), speech rate SR (syll/s), fluency (syll/number of speech chains), mean duration of silences (s), percentage of silence and tonal range (st) were calculated.

Italian Chinese French Russian Italian Chinese French Russian

AR (syll/s)

SR (syll/s)

6.5 5.6 6.1 5.5 7.6 6.1 4.9 5.1

5.6 4.9 4.6 4.7 6.1 5.2 3.8 4.6

Total time Speech Pause (s) (s) 31 36 33 36 26 33 42 39

Fluency (syll/sc)

5 6 11 6 6 6 11 4

14 16 11 13 13 15 10 15

av. duration of pauses (s) 0.402 0.425 0.674 0.409 0.458 0.471 0.604 0.308

Table 11-1: Results of the spectroacoustic analysis. The analysis has shown only one of the considered rhythmic-prosodic parameters is crucially correlated with the persuasiveness perception, since, taken individually, they cannot justify the greater or lesser degree of speakers’ persuasiveness. Pitch contour variation seems to be the unique relevant feature, which combined with the degree of the perceived foreign accent can give account for persuasiveness variation. An initial analysis of the corpus seemed to reveal that the articulation rate was the only feature partly matching the persuasiveness scale, at least for the first three positions. The male Italian, the female Italian and the female French speakers in fact had the highest values, respectively 7.6, 6.5 and 6.1 syllables per second. Nevertheless, taking into account that the Chinese spokesman is the last in the persuasiveness list—though he has exactly the same AR value of the French female speaker—it can be inferred that there should be other factors influencing the persuasiveness processing.

Foreign Accent and Persuasiveness

221

Regarding the speech rate, this parameter gives account for the Italian native speakers, but not for the French spokeswoman: she is the third in terms of persuasiveness, but the last according to the SR list. Fluency and the average duration of pauses do not correlate at all with the results of the perceptual test. As for the tonal range, since it takes into considerations tone variations within each speech chain, but does not take into account the number of times maximum and minimum F0 values are reached, a special index was introduced for this study: tonal variation index (TVI). The TVI is the product of the tonal peaks number obtained through Momel-INTSINT and tonal range in semitones, divided by syllables number (TVI= n. peaks*TR/n. syll). The average TVI values were normalized in order to get inter-speaker comparable data. This calculation, which attempts to provide more accurate data on intonation production, allowed further justifying of the degree of persuasiveness assigned to the French female speaker. The figure 11-5 shows that, just like the male native Italian speaker, the French woman has the lowest TVI values, i.e. she has a flatter, more controlled and less varied intonation. Regarding the female L1 Italian speaker, the TVI value does not match the persuasiveness scale; nonetheless, this does not interfere with the way she is perceived by the informants.

Fig. 11-5: TVI individual values.

222

Chapter Eleven

Since the native Italian man received the best ratings both for the foreign accent and the prosodic production, his intonation pattern can be considered a reference model for the evaluation of all other speakers’ speech. In this perspective, the Italian and the French female speakers can be seen in polar opposition: the first one was evaluated positively in terms of foreign accent (native speaker accent = 96%), but is characterized by an intonational pattern which is different from the persuasive male Italian model and is located, by contrast, on the lower threshold of the foreign speakers’ range. On the contrary, the French woman was recognized as a non-native speaker by all the listeners, although her F0 movement is very similar to that used by the Italian man. Given this, and considering the results on persuasiveness, we can suppose that the French female speaker, using an appropriate intonation, could achieve an acceptable level of persuasiveness (42%) despite her foreign accent. At the same time, the Italian woman, even if recognized as a native speaker, failed to achieve the same level of persuasiveness as the Italian man (female 65%; male 94%) just because of her high TVI value. All this lead us to conclude that both accent and intonation play a role in persuasive communication: however, being recognized as a native speaker has a predominant function, since people usually tend to identify with and, consequently, to prefer familiar voices.

5. Conclusions The data of our research point out that foreign accent, often used in commercials to create a positive mental association with a product thanks to positive stereotyping (i.e. German accent=beer, French accent=perfume, etc.), seems, however, to limit the efficacy of producing persuasion in neutral ads. Native speakers are more persuasive than non-native ones, and listeners seem to seek reassurance from a familiar voice. Persuasiveness in a neutral spot is strengthened also by a flattened and less varied pitch contour. None of the other rhythmic-prosodic features seems to play a relevant role to this regard. Synthesized speech could be a useful mean to verify the relationship between foreign accent and tonal variation with persuasiveness in advertisements not culturally, gender or socially connoted.

CHAPTER TWELVE THE STRATEGIC USE OF PAUSES AND INTONATION IN L2 ITALIAN ARGUMENTATIVE SPEECH: THE CASE OF AN INTERCULTURAL DEBATING BETWEEN CHINESE AND ITALIANS LUISA SALVATI UNIVERSITY OF NAPLES “L’ORIENTALE”, ITALY

1. Introduction Being able to argue in a second language requires not only a good knowledge of appropriate morphosyntactic structures, but also a rich vocabulary and valid arguments to convince the interlocutors. Argumentative skills also need a prosodic component, through which the speaker can convey attention (more or less) to the listener, communicate an inferiority or superiority self-perception with respect to the interlocutor, express emphasis or monotony, certainty or, conversely, uncertainty. These aspects have great importance in argumentative speech, especially if the final task is to persuade an audience, which, indeed, will not only be influenced by what we say but also by how we say it. However, if it is not always easy to manage the L1 prosodic features, what happens if nonnative speakers are asked to debate publicly with native speakers in L2? Starting from this question, this work aims at analyzing the rhythmicprosodic features characterizing the argumentative speech production in L2 Italian compared to L1 Italian.

224

Chapter Twelve

1.1. The debating structure As a research tool for collecting the corpus analyzed in this paper, we have chosen debating, which is not simply a debate where speakers argue about a topic, but rather an interactive exchange of ideas, with a strict protocol of rules managing the alternation of arguments for and against a given topic, imposing a time limit, and finally involving the audience’s judgment to evaluate individual speakers on linguistic, paralinguistic and extralinguistic grounds. When the debate is carried out not only in a L1, but also involves foreigners who argue in L2, then it becomes an intercultural interaction between natives and non-natives, characterized by different linguistic behaviours and cultural backgrounds. In this perspective, the features which normally charactere debating become even more complex because of sociolinguistic and sociopragmatic factors— linked to the different cultural values speakers belong to—and of cognitive factors, related to language learning processes. This type of debating involves intercultural communication skills as well: i.e. proper skills from natives and non-natives to interact by negotiating meanings, values, symbols, ideas, on a “common ground”.1 For this research, a debate was held between a team of Italians and a team of Chinese learners of Italian L2. The debate took place in 2 phases. In the first phase, chaired by a moderator, each team member had up to two minutes to argue, and then he or she was joined by a member of the opposing team. In the second phase, both groups had up to six minutes each to discuss freely, without any moderator, in order to convince an audience.

2. Materials and methods 2.1. The corpus The corpus in Italian L1/L2 was collected via the process of debating, which was audio-recorded by Goldwave 5.58 and videotaped by a Sony HDR-SR8E.2 Furthermore, we made an annotated orthographic transcription

1

Anita Fetzer and Kistin Fischer, eds., Lexical markers of common grounds (Amsterdam: Elsevier, 2006). 2 This corpus is a part of a wider research which also includes a corpus in L1 Italian (referring to a debating among natives speakers) and a corpus in L2 Italian (referring to a debating among Chinese learners of L2 Italian). See Luisa Salvati, “Il debating interculturale. L’interfaccia prosodic-argomentativa nel parlato

Use of Pauses and Intonation in L2 Italian Argumentative Speech

225

of the corpus, based on the CLIPS project—“Lexicons and Corpora of Written and Spoken Italian”. Here we will refer to the spectro-acoustic analysis conducted on the first phase of the debate (see above), where each person spoke without being interrupted. The spectro-acoustic analysis was carried out by Wavesurfer 1.8.1. For each speaker we measured: the number of speech chains; the number of syllables for each speech chain; the duration of each speech chain; the duration of silent pauses, the duration of non-silent pauses, or disfluencies; the maximum and the minimum value of F0 for each speech chain. Furthermore, for each speaker we calculated: articulation rate (AR), that is, the ratio between the number of syllables and the speech chain duration (syll/s); speech rate (SR), that is, the ratio between the number of syllables and the utterance time (syll/s); fluency (F), that is, the ratio between the number of syllables and the number of speech chains (syll/SC), the silence duration percentage; the mean duration of silent pauses(s); the disfluencies duration percentage; the tonal range, defined as the difference between the maximum and minimum value of F0 in an utterance, and calculated in semitones (st), since this measurement allows to compare data relating to different speakers.

2.2. Native and Non-Native participants The Italian speakers were three women and one man, aged between 2025, and they were university students from the Campania region (southern Italy). The Chinese participants were students at the Italian Language and Literature Faculty at the University of Tianjiin. There were two men and two women, between 20-25. At the time of research, they had been living in Naples for four months, and had a language competence in Italian corresponding to B2 level of the Common European Framework of Reference for Languages (CEFR). Before the debate, we separately explained to both groups the rules for debating, and we gave them some tips on how to practice for the debates both individually and in groups. Moreover, a good part of this introductory phase was dedicated to comment on the parameters on which the speakers would be judged: persuasiveness, voice volume, rate, pauses, intonation, posture and gaze, gestures, language use and competence. Subsequently, several debating

argomentativo in Italiano L1/L2” (PhD diss., University of Naples “L’Orientale”, 2012).

226

Chapter Twelve

simulations were held. In the specific case of the Chinese learners, we used a book aimed at developing argumentative skills in Italian L2.3

3. The spectro-acoustic analysis of the intercultural debating Following the procedures described in § 1.1, the two teams debated the motion: “It is proper to use provocative words and images in advertising”. The Italians belonged to the pro-team, while the Chinese belonged to the con-team. The audience consisted of 33 listeners, 20 Italians and 13 Chinese. In the graphs presented in this chapter, speakers have been signed according to the order in which they took turn in arguing (eg. F1 = first speaker), nationality (e.g. IT = Italian), sex (M = male, F = female), type of argument (PRO = in favour, CON = against). The first part of the L1/L2 corpus consists of 4176 syllables, produced in 847 seconds and divided in 322 speech chains. Over an average of 108 seconds, the 4 native speakers produce 35 speech chains, 697 syllables and 12 non-silent pauses (mean values); the non-native speakers, instead, in an average of 72 seconds, perform 46 speech chains, 347 syllables and 21 non-silent pauses (mean values). For an overview of the native and the non-native speech time, see figures 12-1 and 12-2, which show that the non-natives produced exactly twice the silence and disfluencies in percentage. The first prosodic feature to consider is the AR, a rather stable parameter since it is strongly influenced by the anatomical and physiological constraints of the phonatory organs. The AR represents the articulatory gesture quality, and is indicative of hypo-speech or hyperspeech. In hypo-speech, the articulators do not reach the target position for each segment, vowel or consonant: the result is a less accurate speech and a narrower articulatory gesture. Conversely, in a hyper-speech, speech the articulatory movements are produced with greater accuracy and the target positions are always achieved. In this regard, according to the H&H theory,4 speech can be considered as a sort of tightrope: with hypo-speech and hyper-speech, positioned at the opposite ends, the speaker would be 3

Patiz Barki Coricelli and Pierangela Diadori, Pro e contro. Conversare e argomentare in italiano. Livello intermedio. Guida per l’insegnante (Roma: Bonacci, 1997). 4 Byörn Lindblom, “Explaining phonetic variation: a sketch of the H&H theory”, in Speech production modelling, eds. William Harcastle and Alain Marchal (Dordrecht: Kluwer publ., 1990), 403-39.

Use of Pauses and Intonation in L2 Italian Argumentative Speech

227

Fig. 12-1: Percentage of silence.

Fig. 12-2: Percentage of disfluencies.

moving like a tightrope walker, from one end to the other one of the rope depending on factors such as: the speech context, the interlocutor or interlocutors, the conversation topics, the formal or informal style, the types of speech (spontaneous, read, recited, etc.). The more the speaker is

228

Chapter Twelve

able to manoeuvre on the rope, the better the results in reaching communicative purposes. In other words, the AR is closely related to the greater or lesser attention speaker pays to the listener. Hyper-speech is defined as “listener-orientated”, because the speaker uses a low AR, with prolonged syllabic durations, in order to avoid any problem of intelligibility and to be fully understood. On the contrary, a high AR is typical of a “speaker-orientated” speech, where the speaker focuses attention not on the listener but on the self, showing no interest in communicating the message to the interlocutor. In our intercultural debating, the Italian speakers produced an AR of 6.5 syll/s (Fig. 12-3): the data correspond to the mean values indicated for Italian language (5/6 syll/s)5, which reveals a good agility—to pursue Lindblom’s metaphor—by native speakers to move along the tightrope of speech according to the different communication needs.

Fig. 12-3: Articulation rate.

Compared to the Italians, the Chinese produced an AR of 4.9 syll/s. According to Giannini 2000 there are three possible AR degrees: slow, fast, and normal speech. According to her data, the distance between a layer and the next one is 2 syll/s. An acceleration of 2 syll/s would 5

Antonella Giannini, “Range di variabilità della velocità di articolazione in italiano”, in Atti del XXVIII Convegno Nazionale AIA (Trani, 2000), 253-56.

Use of Pauses and Intonation in L2 Italian Argumentative Speech

229

therefore shift from one level to another one. On the other hand, a variation of a 1 syll/s would not be perceived as a level transition, but as a micro-variation within the same utterance.6 It can therefore be said that non-natives, producing a speech with an AR of 2 syll/s slower than the natives, have a “slow” speech, tending to a hyper-articulation. This is probably due not only to the wish to be better understood by the nonnative listeners, but also to the Chinese learners’ difficulty in complying with Italian prosody and producing fluent statements consisting of syllables with a complex morphological structure. As stated by Costamagna, on a production level Chinese learners have difficulty in absorbing the Italian language rhythm and speak, especially in the early stages, word by word, or, rather, syllable by syllable. For this reason in first levels, and, in some cases, also at advanced levels, speech and reading appear very fragmented.7 The second prosodic feature we consider is the SR. Unlike the AR, which is a qualitative parameter, SR can be considered as a quantitative feature, as it expresses the number of syllables actually spoken over a particular period. It thus varies not only according to speech, but also according to silent pauses: the higher the silence, the lower SR, and vice versa. In literature,8 it has been shown that a high SR indicates a dominated position assumed by the speaker, who perceives him- or herself as inferior to the interlocutor and so trying to avoid producing pauses that may give the interlocutor the opportunity to catch the conversational turn. Conversely, a low SR shows a position of “dominance” by the speaker, who feels superior to the listener and is not afraid of being interrupted. Figure 12-4 shows how the native speakers produced a mean SR of 5.4 syll/s, which is within the mean values cited in literature (4/5 syll/s).

6

Ibidem, 256. Lidia Costamagna, “L’apprendimento della fonologia dell’italiano da parte di t,udenti sinofoni: criticità e strategie,” in La didattica dell’italiano a studenti cinesi: il programma Marco Polo ed altre esperienze. Atti del XV seminario AICLU, eds. Elisabetta Bonvino and Stefano Rastelli (Pavia: Pavia University Press, 2011), 52. 8 See among others Danielle Duez, Le pause dans la parole de l’homme politique (Paris: Editions du Centre National de la Recherche Scientifique, 1991). 7

230

Chapter Twelve

Fig. 12-4: Speech rate.

The Italian speakers therefore adopted a speech that was neither too rapid nor very slow, putting the audience in a position of equality. It is plausible that in this case the speech typology and the debate structure played an important role: the produced speech, in fact, not only underlines the impossibility, for native speakers, to self-perceive themselves as superior to the listeners they would have been judged by, but also the futility of self-perceiving in a subordinate position, with little chance of convincing the audience. The mean SR of the four non-native speakers, however, is equal to 3.3 syll/s. These are—whether you consider the mean value, whether it refers to the individual speakers values—very low indices, much lower than the natives’ mean value, with a significant gap of over 2 syll/s. As for fluency, this is a parameter indicating the number of syllables between a silence and the other one. The higher the fluency, the more silent pauses are away from each other. The lower the fluency, the more fragmented the speech. Unlike AR, fluency is a prosodic parameter susceptible to wide variations, depending on both the speaker—as it relates to the ability to vary one’s speech—and the type of speech (prepared, read, spontaneous, etc.). Furthermore, fluency varies depending on other factors: the speaker’s emotional state, and his or her sociocultural background, character, etc. To explain the variations to which

Use of Pauses and Intonation in L2 Italian Argumentative Speech

231

Fig. 12-5: Fluency.

fluency can be subject, we considered the division into three types proposed by Savastano et al.,9 with respect to fluency variability: a first type, concerning the component of emphasis (a lower fluency involves a more emphatic speech), a second type, depending on the degree of control exercised by the speaker (a lower fluency is indicative of a greater control) and, finally, a third type, relative to the speaker’s ability to control his or her speech in real time (the greater the ability, the greater fluency). Comparing our data with the values in literature, the native speakers produce a speech characterized by a fluency which was neither too high nor too low, showing that they had a rather balanced control on their speech, through a moderate use of silent pauses. Conversely, the nonnatives produced a mean fluency of 8 syll: this means that on average they produced about 8 syllables between each silence, an extremely low value compared to the mean native fluency (Fig. 12-5). The data on fluency are confirmed by those relating to the percentage of silence. Figure 12-1 shows that in Italian speakers it is about half that of the Chinese. The silence percentage depends not only on the number of silences, but also on the silent pauses. Regarding the mean duration of silent pauses, we should say that longer silences correspond to a more 9

Elvira Savastano, Antonella Giannini and Massimo Pettorino, “Aspetti prosodici del parlato dei politici”, in Atti del XXIII Congresso Nazionale dell’AIA (Bologna: Editore Cocchi, 1995), 171-76.

232

Chapter Twelve

emphatic speech, compared to a speech with shorter silences. Indeed, silence is one of the most common methods used to emphasize a sample of speech: if, for example, we produce a silence immediately before the word we are going to say, we will obtain the effect of capturing the attention of those who are listening to us; while, to produce a silence, after we have spoken a word, gives the listener time enough to think about what he or she just heard. The mean duration of silent pauses is subject to numerous variations: it depends on the speaker, the type of text, speech, etc. It is very difficult to set a parameter allowing us to determine when a pause can be defined as short or long. In this case, taking into account that we have evaluated as silent pauses all pauses longer than 0.1 s, we have labelled as short those pauses that are within a range between 0.1 and 0.3 s.; as medium those between 0.4 and 0.6 seconds; and as long those pauses exceeding 0.6 s. Starting from this premise, our four native speakers produced average pauses of 0.4 seconds. As we can observe, the four native speakers did not put too much emphasis on their sentences, producing more jointing pauses—to indicate syntactic transition moments—rather than emphatic pauses. The mean value reached by the non-natives was 0.5 s (Fig. 12-6). The empty pauses made by the non-natives were slightly longer, lasting 0.1 second more: the nature of their silence was not emphatic, but they are useful to control utterances, in terms of content and form.

Fig. 12-6: Mean duration of silent pauses.

Use of Pauses and Intonation in L2 Italian Argumentative Speech

233

Another prosodic element consists of non-silent pauses or disfluencies.10 In most cases, they are unintentional pauses, and are considered as moments of planning—when the speaker must make choices—or as a signal the speaker gives to the interlocutor in order to maintain the conversational turn. Disfluencies are also signs of insecurity or hesitation on the part of by the speaker, who fills the time between a word and the following one by producing non-silent pauses. In Fig. 12-2 the non-native speakers produce about twice as many disfluencies (10.6%) as the natives (5.7%).

Fig. 12-7: Tonal range.

The last prosodic parameter analysed regards the tonal range, one of the most important prosodic feature in terms of pragmatic efficiency, as it relates to different levels of communicative interaction. It varies depending on the speaker (age, sex, socio-cultural status and emotional state), the speech contextual situation (formal, informal, relationship between speakers), the speech typology (spontaneous, recited, semispontaneous, read, etc.) and topic. In general, a restricted tonal range is indicative of a monotone and flat speech, while a wide tonal range shows a varied and dynamic speech.

10 In this research we consider as disfluencies all vocalizations, nasalizations, extensions, auto-repetitions and auto-corrections.

234

Chapter Twelve

In Fig. 12-7 the native speakers TR, if compared with literature data for Italian L1, is on average quite large, varied, expressive, addressed to the interlocutor. In this sense, the native speaker produces a speech adequate to the task of convincing an audience. With the non-natives TR, it has a mean value of 6.4 semitones. The significant difference between the pitch movement in L2 and L1 is indicative of the Chinese learners’ difficulty not only in managing the variations in intonation—to discern an exclamation, from a question or a statement—but also in perceiving the pragmatic value behind the pitch, which allows a single sentence to take multiple functions and meanings in communication. In Italian, in fact, syllables assume different heights allowing, for example, for the distinction between an affirmative sentence (pattern plate and subsequently descending) and an interrogative one (pattern characterized by tonal peaks, ascending/descending, not necessarily at the end of the sentence). The tonal variation can be extended to all the syllables of the utterance, assuming a distinctive value in prosodic terms. In Chinese, instead, the high tone is a distinctive feature on the semantic level, distinguishing words with different meanings. As explained by Xianonan,11 in Chinese the tone is assigned to each syllable—stressed and unstressed—and the intonation of a sentence can end with intonation contours of different heights, depending on the tonal contour of the final syllable. Finally, the intonation form (ascending, descending or suspension) does not have a distinctive value. It is thus possible to explain why the non-natives produce a speech tending to be flat and monotone, compared to the dynamic argumentative speech of the natives.

4. Evaluations of audience The final outcome of the intercultural debating is quite interesting because of the heterogeneity between the evaluations assigned to the single speakers, and the final vote expressed by the audience. Let us see it in detail. After the debate, when the audience were asked to choose the team that persuaded more, the con-team (the Chinese speakers) won, with 30 votes out of 33 (in line with the individual assessment on persuasiveness). In the evaluation given by the entire audience, the final vote contrasts with the ratings given to the individual speakers because it 11

Shen Susan Xianon, “Ability of Learning Prosody of an Intonational Language by Speakers of a Tonal Language: Chinese Speakers Learning French Prosody”, IRAL 2 (1990): 119-34.

Use of Pauses and Intonation in L2 Italian Argumentative Speech

235

seems to favour the pro-team (the Italian speakers). The only exceptions are three parameters: rate, language use and competence, and persuasiveness, in which the Chinese team won. This leads us to think that, on the one hand, the overall assessment given at the end of the debate was influenced by the cooperative work (and obviously by the expressed arguments), and on the other, that the parameters related to the prosodic dimension (rate), and linked to the pragmatic dimension (language use and competence), are higher in the Chinese team, which ultimately proved more convincing. In our opinion, these two important language components played an important role in the argumentative speech production and is worthy of further investigation on the relationship between prosody, pragmatics and persuasiveness.

5. Conclusions In conclusion, in our study when native and non-native speakers debated before native and non-native listeners, the Chinese speakers produced an hyper-articulated speech, compared to the Italian (respectively: AR= 4.9 syll/s vs. 6.5 syll/s; SR = 3.3 syll/s vs. 5.4 syll/s). Furthermore, the Chinese speech was interrupted by numerous silences (fluency = 8.3 syll), compared to the Italian fluent speech (fluency = 20.4 syll): the mean duration of silent pauses was neither too short nor very long in both groups, showing a gap between them of 0.1 s (0.5 s = Chinese vs. Italian = 0.5 s). The Chinese speech was also fragmented by a significant percentage of disfluencies, about twice of that produced by the natives (5.7% vs. 10.6%). Finally, the non-natives did not show a tonal change in their speech compared to the more dynamic discourse of the Italians (6.4 vs. 10.8 st). These data and the stability emerged from a comparison with the other corpora of the research12 collected shed new lights on the studies of prosodic production in L1 and L2. On the one hand, the constancy of the numerical data legitimate their use in literature as reference values for the semi-spontaneous and spontaneous speech, on which there are few studies and little empirical research, especially as regards the prosodic production in L2 Italian. On the other hand, in the second language acquisition study, the stability of the results opens the way to the possibility of creating, as previously mentioned, a model for the suprasegmental level of interlanguage and allows, in education, for more precise assessment of the learner’s prosodic skills. In this sense, we hopefully would extend this research to other language competence levels 12

See note 2.

236

Chapter Twelve

of the Common European Framework of Reference for Languages and undertake a comparative analysis of the different stages of language learning.

CHAPTER THIRTEEN PROSODIC-PRAGMATIC INTERFACE IN L2 ITALIAN LEARNERS’ REPETITIONS ANNA DE MARCO UNIVERSITY OF CALABRIA, ITALY

AND PATRIZIA SORIANELLO UNIVERSITY OF BARI, ITALY

1. Introduction One of the ways through which it is possible to define repetition is by resaying units of talk or producing two utterances that have the same propositions.1 As Schegloff2 puts it, “There are several places in conversation that is, either sequential contexts or specific practices of talking in which speakers seem demonstrably orientated to producing talk that ‘says the same thing’ as was said before and does so by saying it ‘in the same words’”. Looking at a recent definition by Tannen,3 repetition can be defined as the way meaning is created by the recurrent and a different contextualization of words and phrases in discourse. There are several places in conversation in which speakers seem orientated to

Although the paper is the result of the joint work of the two authors, De Marco is responsible of the paragraphs: 1., 2., 2.1., 5.2., 5.3., 6, and Sorianello for the paragraphs: 3, 5.1., 5.1.1., 5.1.2., 5.1.3., 5.1.4; the § 4 is common. 1 Harvey Sacks, Lectures on Conversation, vol. 1 (Oxford: Blackwell, 1992); Jefferson Gail, “On the poetics of ordinary talk”, Text and Performance Quarterly 16 (1996): 1–61. 2 Emmanuel A. Schegloff, “On dispensability”, Research on Language and Social Interaction 37/2 (2004): 95–149. 3 Deborah Tannen, Talking voices: repetition, dialogue and Imagery in Conversational Discourse, 2nd edition (New York: Cambridge University Press, 2007).

238

Chapter Thirteen

produce talk that “says the same thing” as was said before and does so by saying it “in the same words”.4 The linguists who have been investigating on conversational analysis have tried to identify schemas and patterns of interaction particularly in the area of the sequential organization of turns. The outcomes of the interaction analyses stand out as a very complex picture of communicative practices that are extremely sensible to the realization of the input instant by instant by the participants engaged in locally managed conversation and interaction.5

2. The pragmatic aspect of repetitions In order to explain the meaning of any instance of repetition it is important to refer to its basic categories: 1) who repeats; 2) what is repeated; 3) where the repetition occurs; and 4) what the repetition does. In our study, we will particularly refer to repetition produced by the nonnative speaker who is, in most cases, the recipient of the repeats. In other cases it is also interesting to see how repeats are played by the two interactants, i.e. produced by the speaker of the source utterance and by the recipient, self- and other-repetition. Repetition has been found to act in different sequential intersection: in addition, to gain a turn,6 to halt a sequence-in-progress,7 and to close a sequence.8 In the case of a repair sequence, repetition can have the function of initiating a repair on a previous speaker’s talk9 and performing or accomplishing self- or other-initiated self-repair.10 Repetition is also 4

Schegloff, “On dispensability”, 119–20. Cecilia E. Ford and Sandra A. Thompson, “Interactional units in conversation: Syntactic, intonational, and pragmatic resources for the management of turns”, in Interaction and Grammar, eds. Elinor Ochs, Emanuel A. Schegloff and Sandra A. Thompson (Cambridge: Cambridge University Press, 1996), 134–84. 6 Schegloff, “On dispensability”, 21-30; Charles Goodwin, “Restarts, pauses, and the achievement of mutual gaze at turn-beginning”, Sociological Inquiry 50/3–4 (1980): 272–302. 7 Tanya Stivers, “No no no and other types of multiple sayings in social interaction”, Human Communication Research 30/2 (2004): 260–93. 8 Traci S. Curl, John Local and Gareth Walker, “Repetition and the prosodypragmatics interface”, Journal of Pragmatics 38 (2006): 1721-51. 9 Emanuel A. Schegloff, “Third Turn Repair”, in Towards a Social Science of Language. Social Interaction and Discourse Structures, Papers in honor of William Labov, Vol. 2, eds. Gregory R. Guy et al. (Amsterdam: John Benjamins, 1997), 31-40. 10 Traci S. Curl, “Repetition’ repairs: The relationship of phonetic structure and sequence organization”, in Sound Patterns in Interaction: Cross-Linguistic Studies 5

Prosodic-Pragmatic Interface in L2 Italian Learners’ Repetitions

239

employed to confirm a request,11 or to continue participation in conversation “even though one has nothing new to add”.12 Wong suggests that “if repetition is a relevant part of our language system and behavior, then there must be some orderliness to it and regularity about it”.13. Schegloff14 maintains that participants may repeat a turn beginning in order to overcome possible trouble at the point of speaker change. Schegloff15 argues that speakers may find their utterances to have been ineffective (i.e. recipients have not displayed any responsiveness to the source utterance) and, by repeating, they may try to reinstate the sequential implicativeness of their utterances. Self-repetition is also used for organizing larger conversational units, such as story-telling sequences. Tannen16 differentiates between two types of repetitions, synchronic and diachronic: the first deals with repetitions occurring in the same turn and the second with repetition occurring after some time. In terms of sequential organization of function of repetition we can have two main types of repetition, i.e. self-repetition and “allo-repetition”. The first is the immediate repetition of what the speaker has said immediately before (i.e. the speaker repeats himself). Self-repetition can also include interference by the other speaker. Bazzanella17 calls this last type monologic repetition, i.e. when the speaker wants to signal uncertainty or to emphasize a refusal. With “allo-repetition”, one refers to the repetition of what another speaker has said which can go across several turns and therefore involves

from Conversation, eds. Elizabeth Couper-Kuhlen and Cecilia E. Ford (Amsterdam: John Benjamins 2004), 273–98; Traci S. Curl, “Practices in otherinitiated repair resolution: the phonetic differentiation of repetitions”, Discourse Processes 39 (2005): 1–44. 11 Marja-Leena Sorjonen, “Vuoronalkuiset konnektorit: mutta [Turn-initial connectors: mutta ‘but’]”, in Kieli 4: Suomalaisen keskustelun keinoja I [Characteristics of Finnish conversation I], ed. Auli Hakulinen (Helsinki: Department of Finnish, University of Helsinki, 1989): 162–76. 12 Tannen, Talking voices, 69. 13 Jean Wong, “Repetition in conversation: A look at “first and second sayings”, Research on Language and Social Interaction 33/4 (2000): 420. 14 Schegloff, “Third Turn Repair”, 31-40. 15 Schegloff, “On dispensability”, 122-25. 16 Tannen, Talking voices, 2. 17 Carla Bazzanella, “Forme di ripetizione e processi di comprensione nella conversazione”, in La conversazione. Un’introduzione allo studio dell’interazione verbale, eds. Renata Galatolo and Gabriele Pallotti (Milano: Cortina, 1999), 20525.

240

Chapter Thirteen

both parties.18 One of the examples of allo-repetition is polyphonic repetition when fixed structures, based on conversational routines—such as greetings—are retrieved after some turns. As for the item which is repeated, repetition can involve the same exact item or a paraphrase of it. In the middle of these two poles we find a variation such as a question transformed in a statement, a change in the categories of time and person. In relation to the point in which the repeated item appears, selfrepetitions can be found: a. At the beginning to gain the turn (in case of an overlap); b. In the same turn in order to keep it or to gain attention or to maintain it during a simultaneous discourse; c. At the end to give a turn; to retrieve a turn in a distance after a prolonged interruption19. Among the other functions and acts that hetero-repetition generates, we find textual function: to reply to a question, to answer in a routine way with the same formula (see above), to imitate, to cite, to comment. Other further functions Bazzanella adds to these are situated on a scale of agreement/disagreement where grade zero corresponds to the acquisition of information, then agreement to what the other speaker says, and disagreement correlated to the correction in the form of negative feedback.20

2.1. Repetitions in native/ non-native speakers interactions The pragmatic analysis of repetition has not received very much attention in the interactions of L2 speakers. Investigations have been orientated towards the observation of the relationship between use of repetition and learning of a second language and the pedagogical use of repetition in L2 classroom as an opportunity for the students to manipulate grammatical structures. Following Ciliberti,21 repetition occupies an important role as an explicit strategy and as an indication of the adoption of other strategies of 18

Xiaoling Zhang, “Echoing in Real-life English Conversation”, PALA: The Poetics and Linguistics Association, Occasional paper n. 9, 1998, 1-16. 19 Bazzanella, “Forme di ripetizione e processi di comprensione nella conversazione”, 205-25. 20 Miriam Voghera, Sintassi e intonazione nell’italiano parlato (Bologna: Il Mulino,1992). 21 Anna Ciliberti, “Repetition in native/non-native interaction”, in Repetition in dialogue, ed. Carla Bazzanella (Tübingen: Niemeyer, 1996), 39-49.

Prosodic-Pragmatic Interface in L2 Italian Learners’ Repetitions

241

facilitation in native/non-native (N/NN) interactions. Ciliberti envisages different functions expressed by the repetition both by the native and the non-native speakers. As for the non-native speakers she signals the following functions: allowing speakers to contribute to conversation without effort by simply replaying the speaker’s preceding utterance; a way to practice linguistic forms; a control over acquisition; a way to signal rapport and active listening. The native speaker’s repetition is a way to correct the interlocutor’s production in marginally signalling an active listening role and rapport. The relation between native and non-native speakers is certainly a very particular type of interaction if we consider, as Ciliberti underlines, that social perceptions of their reciprocal roles and status may influence their performance. The most crucial characteristic of Ns/NNs interaction is the interlocutors’ awareness of their difference, their native non-native speakerness. Looking at conversations between native and non-native speakers of English, Wong22 identifies a type of selfrepetition that is used exclusively by native speakers in her data: repetition of a lexical item, clause or sentence within the same turn in order to resume story telling after an insertion of some parenthetical information, the second part being lower in pitch and shorter in duration. Two important outcomes of the research on non-native/native interactions are, first, that the function of repetition in non-native speakers is more orientated towards comprehension23 intended as a control on both form and function of the repeated lexical item or expression. Second, repetition is linked to the level of proficiency of the learner: the more competent the learner, the greater the involvement in the conversational moves, and the more specialized are the strategies involved in the use of repetitions. With our contribution, we would like to offer a general overview of the different communicative functions of repetition in a corpus of non-native speakers of Italian as they are employed in a semi-structured interview with a native speaker. We want to underline the overall range of possibility to use repetition as a linguistic device to meet the communicative need in a particular type of interaction. In order to clear up the various functions used by the speakers, we have investigated the relationship between communicative functions and prosody to help us to disambiguate some of the functions emerging in the 22 Jean Wong, “Repetition in conversation: A look at ‘first and second sayings’”, 407–24. 23 Eva Wiberg, “Il riferimento temporale nel dialogo. Un confronto tra giovani bilingui italo-svedesi e giovani monolingui romani”, Études romanes de Lund 58 (1997): 318.

242

Chapter Thirteen

corpora. Since the phenomenon of repetition is not a unitary phenomenon we realize that the possible range of meanings, functions and use of repetition are not necessarily the same for all speakers, and also that they cannot be all captured by the analysis of prosody. The pragmatic function of prosody has been the focus of a growing interest in the recent literature on the pragmatic/prosody interface. Nevertheless, systematic studies on the role of prosody in the discursive practices are very few. Even fewer are contributions delineating the precise nature of this interaction in the L1 and L2. Looking at prosody from the interactional perspective is useful, not just to match a certain prosodic form to a precise semantic, grammatical or pragmatic meaning but, most of all, to observe how it is used, i.e. what kind of interactional or conversational functions it produces through the orientation of participants towards their prosodic resources. Our main aim is to start from the analysis of some of the prosodic resources, together with some other elements such as lexis or articulatory details among others, accompanying the production of repetition in order to develop a taxonomy of possible meanings in terms of conversational and interactional moves or practice, locally or more globally referred to by the speakers engaged in an interaction in Italian as L2. Since our corpora come from speakers with different levels of proficiency, we have divided them up in one group (A2 proficiency level), and a control speaker (B2 proficiency level) in order to look into the different strategies adopted related to different degrees of competence in the L2.

3. Prosodic aspects of repetitions We now turn to the issue of prosodic realization of repetitions. Spontaneous speech shows significant rates of disfluencies (about one every 10-20 words) such as: filled pauses, repetitions, false starts, etc.24 Among these, repetition represents the most frequent disfluency occurring in spontaneous and dialogical speech, with a percentage rate of about 21.9%.25 In recent years, the studies on repetition drew the attention of 24 According to Elizabeth E. Shriberg, “To errrr is human: ecology and acoustics of speech disfluencies”, Journal of International Phonetic Association 31 (2001): 160, disfluency types comprehend the following phenomena: filled pauses, repetition, deletion, substitution, insertion and articulation error. 25 Jennifer Cole et al., “Prosodic parallelism as a cue to repetition disfluency”, in Proceedings of DISS ‘05, Disfluency in spontaneous speech workshop, Aix-enProvence (2005), 1-4.

Prosodic-Pragmatic Interface in L2 Italian Learners’ Repetitions

243

many scholars in more theoretical perspectives, as well as in different applied fields. However, this research area raises many problematic issues; first of all, it should be noted that both the terminology and the methods used display a relevant heterogeneity. Also, the experimental findings are only partially comparable. This is not surprising, since the empirical protocols that inspired much of this field research are quite different. For a long time, the phonetic aspects of repairs26 have been a privileged psycholinguistic topic. Early surveys aimed to define the cognitive architecture of speech. These studies27 focused on verifying the timing of repairs, the strategies of speech replanning, lexical truncation, error repairs and true repetitions of words.28 In prosodic perspective much information about repetitions is available in the phonetic studies on verbal disfluencies. More recently, repetitions, and on a wider scale, self-repairs, were studied most extensively against a phonetic background. Generally, repetitions are marked by specific prosodic cues such as temporal duration, F0 contour, change in intensity, pauses and glottalization. The activation of one or more phonetic features contributes to signalling the repetition to the listener. For instance, an increase of F0 values in correspondence of altered or corrected words represents a very common repair strategy. The importance of these phonetic parameters has also been verified also by means of perception tests, or in statistical production models. With respect to that, Nakatani and Hirschberg29 elaborated the Repair Interval Model (RIM), a computational and statistical model of repair site detection based on classification and regression tree techniques. According to Hindle,30 Levelt31 and Bear et al.,32 all disfluencies—and among these, repetitions—are acoustically marked by a special “edit 26

Generally speaking, repairs are defined as the self-correction of a word or sequence of words within an utterance. 27 Sieb G. Nooteboom, “Speaking and unspeaking: detection and correction of phonological and lexical errors in spontaneous speech”, in Errors in linguistic performance, ed. Victoria Fromkin (New York: Academic Press, 1980), 287-305; Willem Levelt, “Monitoring and self-repair in speech”, Cognition 14 (1983): 41104. 28 For this purpose, it is relevant to observe that psycholinguistic research contains few prosodic information having been for the most part test-based. 29 Christine H. Nakatani and Julia Hirschberg, “A corpus-based study of repair cues in spontaneous speech”, Journal of Acoustical Society of America 95 (2009), 1603-16. 30 Donald Hindle, “Deterministic parsing of syntactic non-fluencies”, in Proceedings of the 21st annual meeting of the association for computational linguistics (1983), 123-28. 31 Levelt, “Monitoring and self-repair in speech”, 303-42.

Chapter Thirteen

244

signal”, a perceptually salient phonetic structure marking the point in which happens the interruption of speech flow. Repetitions have an internal structure formed by the following threeregions. 1)

Reparandum (R1): it is the verbal portion that will be later corrected or replaced. The offset constitutes the interruption point, namely the site at which disfluency started. 2) Editing Phase: it is an intermediate structural unit, a segmental space whose extension spreads from the interruption point to the onset of repair. It may be empty or occupied by a silent or filled pause (i.e. I mean ehm, uhm, etc.). This phase is very relevant to the functioningof systems based on automatic speech segmentation, since it improves the precision rate of repair site detection. 3) Repair (R2): it includes the lexical material which has to be repaired. It is also the point at which the resumption of normal condition of verbal fluency occurs. The three-regions structure is displayed below in Fig. 13-1.

Fig. 13-1: Regions in a disfluency, from Shriberg33.

The three units are linearly ordered, because they have a rigid syntagmatic surface structure. Several studies have shown that the removal of the first two regions (i.e. reparandum and editing phase) is able to restore a fluent version of the utterance. The three-regions structure is distinctively characterized by particular phonetic manifestations. It is important to note that major phonetic effects are concentrated near or around the “interruption point”. As regards temporal behaviour, the final syllables of the reparandum are normally lengthened; if the reparandum contains a word cutoff (20-25% for Cole et 32 John Bear, John Dowding and Elizabeth E. Shriberg, “Integrating multiple knowledge sources for detection and correction of repairs in human-computer dialog”, in Proceedings of the 30th Annual meeting of the association for computational linguistics (1992), 56-63. 33 Shriberg, “To errrr is human”, 160.

Prosodic-Pragmatic Interface in L2 Italian Learners’ Repetitions

245

al.34), laryngealization and alteration of voice quality are very frequent. Reparandum generally shows a longer temporal duration than in repair (R2); it in fact contains that typical articulatory lengthening which anticipates the verbal interruption flow. In addition, R2 has a more limited tonal excursion as compared with R1. Most of the time, both R1 and R2 should have a falling intonational pattern; but when a particular emphasis is realized on R1, the final movement, on the contrary, is rising. Repetition is a common phenomenon both in native and in non-native language. Nevertheless, there are not many studies devoted to the exploration of this disfluency phenomenon in second language acquisition. It is important to observe that the language classroom is the privileged setting of the studies carried out in this area.35 According to the findings collected in most studies, repetition in non-native speakers seems to be a more frequent communicative strategy than in native language.36 The prosodic aspects of sentences containing repetitions have been primarily observed in native speech.37 Conversely, the literature available on non-

34

Cole et al., “Prosodic parallelism as a cue to repetition disfluency”, 1-4. Roy Lyster, “Recasts, repetition, and ambiguity in L2 classroom discourse”, Studies in second language acquisition 20 (1998): 51-81; Gláucia Silva and Denise Santos, “Framing participation through repetition: the case of a Portuguese learner in different settings”, Portuguese Language Journal 1 (2006): 1-24; Bada, Erdo÷an, “Repetitions as vocalized fillers and self-repairs in English and French interlanguages, Journal of pragmatics 42/6 (2010): 1680-88; Bilal Genç, Mustafa Mavaúo÷lu and Erdo÷an Bada, “Types and functions of repetitions in the narrations of Turkish speakers of French”, Research on Youth and Language 4 (2010): 216-24; Lyster, “Recasts, repetition, and ambiguity in L2 classroom discourse”, 51-81. 36 Stewart Stuart and Lynn Pearson, “Development of communication strategies among foreign language learners”, Mountain interstate foreign language review 5 (1995): 122-27; Zoltan Dörnyei and Mary Lee Scott, “Communication Strategies in a Second Language: Definitions and Taxonomies”, Language Learning 47/1 (1997): 173-210. 37 Willem Levelt and Anne Cutler, “Prosodic marking in speech repair”, Journal of Semantics 2 (1983): 205-17; Judih Kormos, “Monitoring and Self-repair in L2”, Language learning 49/2 (1999): 303-42; Madeleine C. Plauché and Elizabeth E. Shriberg, “Data-driven subclassification of disfluent repetitions based on prosodic features”, in Proceedings of the International Congress of Phonetic Sciences, San Francisco, USA, 1999, eds. John J.Ohala et al. (Berkeley, University of California, 1999), vol. 2, 1513-16; Elizabeth E. Shriberg, “Phonetic consequences of speech disfluency”, in Proceedings of the International Congress of Phonetic Sciences, San Francisco, USA, 1999, eds. John J. Ohala et al. (Berkeley, University of California, 1999), vol. 1, 619-22; Shriberg, “To errrr is human”, 153-69; Cole et 35

246

Chapter Thirteen

native speech does not provide adequate prosodic information about repetitions. As a consequence, most of the aspects of this research field are mainly unexplored. For the Italian language, short impressionistic observations are available only in Bazzanella and Pallotti.38 In particular, Pallotti (2005) focused on the self-repetitions realized by a Moroccan fiveyear old during her first exposure to Italian language. From a prosodic point of view, the author observed that repetitions displayed a continuation contour and a fast speech rate, excepting the last word of the utterance which was normally lengthened.

4. Methodology Six beginners of Italian L2 were examined (proficiency level A2). They were all young and female (aged between 21-26), four were Turkish native speakers while the others were Spanish ones. Two different strategies of speech elicitation were employed. Firstly, all the participants were interviewed about socio-cultural differences between Italy and their native country in order to obtain a series of semi-spontaneous dialogical conversations. Secondly, we asked to them to describe sequences of images, appositely chosen for the research. An American informant (aged 30) with a better competence of Italian as second language (proficiency level B2) was considered in order to test whether the frequency and typology of repetitions was conditioned in same way, and to what extent to a higher degree of proficiency level of Italian language. Her data were elicited through an interview on the subject of her family history and personal experiences in Italy since her parents migrated to the United States when she was a baby. The prosodic analysis of her speech was not fully completed. We found her speech extremely interesting, and more complex than that of the other less proficient speakers, and sometimes we found it hard to capture the sameness of what at a first glance could appear the same phenomenon, i.e. repetition. We therefore attempted to view the single utterances containing the repetition in a more dynamic way taking into account the whole organization of the conversational turns. al., “Prosodic parallelism as a cue to repetition disfluency”, 1-4; Curl et al., “Repetition and the prosody-pragmatics interface,” 1721-51. 38 Carla Bazzanella, “Aspetti pragmatici della ripetizione dialogica”, in La linguistica pragmatica. Atti del XXIV Congresso della SLI, ed. Giovanni Gobber (Roma: Bulzoni, 1992), 433-54; Bazzanella, “Forme di ripetizione e processi di comprensione nella conversazione”, 205-25; Gabriele Pallotti, “Variations situationelles dans la construction des énoncés en L2: le cas des autorépétitions”, AILE - Acquisition et interaction en langue étrangère 22 (2005): 101-30.

Prosodic-Pragmatic Interface in L2 Italian Learners’ Repetitions

247

All the speakers were digitally recorded in a quiet room by means of a Tascam DR-07 (sampled at 22050 Hz, 32 bits); the linguistic material was orthographically transcribed. It was subsequently segmented in conversational units and acoustically analyzed by means of PRAAT. In this study, we will refer to repetitions according to the theoretical framework adapted from Levelt.39 The findings on the acoustic and prosodic properties of self-repetitions in non-native Italians were discussed. We investigated the possibility of matching the pragmatic point of view with the prosodic one. To this end, we considered those prosodic parameters that normally occur in the vicinity of the disfluency interval, as summarized below: .

(i) The length of unfilled and filled pauses; (ii) R1 and R2 temporal duration; (iii) R1 and R2 intonation contours and pitch excursion.

5. Discussion of the results 5.1. Group A2 We recognized various types of repetitions in Turkish and Spanish learners. At this stage of the analysis, significant differences between the two groups of learners did not emerge. This suggests that repetition is a general communicative strategy strongly connected not only with the L1 language, but also with the L2 proficiency. As already mentioned, the terminology, as well as the range of functions conveyed by repetitions, are not always comparable. For instance, Silva and Santos40 distinguished several functions for L2, among which were: stalling, practice, participatory listenership, ratification of something previously said and meta-repetition; in addition, they noted that the repertoire of functions is wider in the interview setting than in the classroom. Diversely, Genç et al.41 distinguished only two categories: filler type and self-repair, a sort of monitoring form. There is a wide range of functions covered by self-repetition in our speech sample. The pragmatic categories are reported below, listed according to their statistical incidence: 39

Levelt, “Monitoring and self-repair in speech”, 41-104. Silva and Santos, “Framing participation through repetition”, 1-24. 41 Genç, Mavaúo÷lu and Bada, “Types and functions of repetitions in the narrations of Turkish speakers of French”, 216-24. 40

248

Chapter Thirteen

1) Self-repair: this type of repetition implies always a correction of one or more sequential words. It works as a true self-correction of an accidental lapse, a problem with the availability of lexical forms or grammatical structures. 2) Stalling: it reveals a strong planning difficulty. Very often the sentence contains also word cut-off, a notable amount of silent and filled pauses both before and after the repetition. In non-native language such a repetition is a clear sign attesting the lack of language competence. Typically, they occur in a difficult interactional context. 3) Expanding: this repetition represents the starting point of a reformulation of a part of the utterance. In a certain sense, it helps the speaker to develop and continue the conversation. 4) Emphasis: this repetition gives a certain prominence to the repeated words. The utterance does not show either reformulation or correction, since this category represents a perfect repetition. Normally, the two lexical repeated sequences are contiguous and without any pause interruption. 5) Control over form or meaning: by means of such form of repetition, the learners are asking for agreement about the correctness and adequacy of the repeated word. In Table 13-1 we reported the percentage rate of each discussed category for Turkish and Spanish learners. Nevertheless, during the analysis we met some theoretical difficulties concerning the classification of repetition. The assignment of a repetition within a precise pragmatic category is not a simple or automatic procedure: sometimes, the repetition shows a complex pragmatic structure. For instance, we found recurrences of stalling repetitions also containing self-repairs (ex. ti odiano/ti odia ti odio) or self-repair repetition that suggests an interruption in the speach. Function Self-Repair Stalling Expanding Emphasis Control over form/meaning

% 32 29 17 14 8

Table 13-1: Percent rates of repetition types.

Prosodic-Pragmatic Interface in L2 Italian Learners’ Repetitions

249

In the following sections we discussed the acoustic results concerning pause length (§ 5.1.1.), R1 and R2 duration (§ 5.1.2.) and intonation contours (§ 5.1.3.). 5.1.1. The pauses In L2, there are a high number of silent and unfilled pauses. Contrary to what was expected, in non-native Italian the interruption of normal flow of speech does not coincide with the end of reparandum. Not only does each conversational turn contain a pause, but the pauses are also frequently produced before and after the repair, determining a new interruption point and the necessity of a new lexical reparation. The behaviour is recursive, since the repair changes its pragmatic status, beginning a new reparandum. As far as duration is concerned, we noted that silent pauses produced before or after reparandum do not show a great temporal difference (623 ms vs. 591 ms); the same tendency was observed for pauses preceding or following the repair (548 ms vs. 587 ms). On the contrary, filled pauses were longer when they recurred before the reparandum, but not before the repair. The presence of a pause preceding R1 suggested that the learner has a trouble planning. The average pauses length is illustrated in Table 13-2.

Reparandum Edit phase Repair

Preceding Silent pauses 623 (321)

Following Silentpauses 591 (340)

Preceding Filled pauses 917 (303)

Following Filled pauses 491 (239)

548 (257)

587 (180)

700 (102)*

808 (274)

Table 13-2: Mean duration (in ms) of silent and filled pauses, both preceding and following R1 and R2. In brackets the Standard Deviation values. 5.1.2. Duration R1 and R2 did not display the same duration. Interestingly, their length depends on the pragmatic function conveyed by the repetition. As shown in Table 13-3, R1 systematically has a longer duration with respect to R2, with the only exception of stalling type in which R1 and R2 are practically equal (i.e. 551 ms vs. 523 ms). The greatest difference is in the control over form function (i.e. 762 ms vs. 355 ms) followed by emphasis type (i.e. 815 ms vs. 647 ms). Temporal lengthening in R1 is often associated

Chapter Thirteen

250

with alterations in voice quality such as creaky voice or glottalization. Paying more attention, we note that in stalling type, R1 has a lower duration, while in emphasis and control over form types, R1 shows higher duration.

R1 R2

SelfRepair

Stalling

Expanding

Emphasis

686 544

551 523

673 506

815 647

Control over form 762 355

Table 13-3: Mean duration (in ms) of R1 and R2 for repetition types. These results coincide only in part with data available in literature. For instance, according to Shriberg42 and Curl et al’s43 data, reparandum shows a remarkable length if compared to repair; on the other hand repair has the same duration of the fluent version of the repeated word. Nevertheless, Shriberg says that “not all repetitions show the same pattern”. The cause is probably due to the typology of the speech corpus under investigation: native speech in Shriberg, as in most of studies devoted to this topic; non-native in our research. 5.1.3. Intonational contours We distinguished five F0 contours. In detail, (1) falling: a decreasing F0 movement, (2) high-fall: the falling starts from a very high F0 value, (3) level: a flat and monotonous F0 pattern, (4) rising: an increasing F0 movement, (5) low-rise: it is a subtype of the rising pattern, but in this case the rise starts from a low F0 value.44 In Table 13-4 we reported the percentage rates of each intonational pattern observed for each repetition class. As with duration, intonational contours also seem to be strictly related to the repetition types. Furthermore, the F0 contours differ according to 42

Shriberg, “Phonetic consequences of speech disfluency”, 619-22; Shriberg, “To errrr is human”, 153-69. 43 Curl et al. “Repetition and the prosody-pragmatics interface”, 1721-51; Shriberg, “To errrr is human”, 153-69. 44 We distinguished high-fall and low-rise from their counterparts falling and rising on the basis of pitch range. To be precise, we considered high-fall and low-rise only when the tonal excursion of the F0 movement was equal or higher of 5 semitones (ST).

Prosodic-Pragmatic Interface in L2 Italian Learners’ Repetitions

251

the structural position of the repeated words. Some F0 patterns are in fact prevalent in certain repetition classes. For instance, control over form displays, in both R1 and R2, only high patterns. On the contrary, stalling type shows only low contours. The other typologies are more varied, as discussed below.

F0 Contour Falling High-Fall Level Rising Low-Rise

SelfRepair R1 R2

Stalling

Expanding

Emphasis

R1

R2

R1

R2

R1

R2

Control over form R1 R2

40 2 2 55 1

85 0 15 0 0

38 0 62 0 0

17 3 5 7 65

80 2 0 18 0

0 12 0 3 85

10 72 8 0 10

0 20 0 32 48

30 4 10 56 1

0 0 0 86 14

Table 13-4: Percent rates of F0 contours. 5.1.4. Discussion In this section we report the prosodic features of the typologies observed, pointing out the prevalent configuration of the most representative repetitions. There is a positive correlation between the melodic contour of R1 and R2 and the pragmatic function of repetition. In order to define these aspects, and find some systematic behaviour, we considered various classification parameters. The most frequent typology observed in our learners’ corpora is Self-Repair repetition (32%), cf. Table 13-5, see the examples (1-3) (1) ma re+risolto risa+risolto di l’esame 45 (2) perché venuta è venuta sono venuta qui quindi dovia parlare (3) Oggi in Turchia >ehm> c’è ne+ nevi nevi

45

We used the following transcription conventions: : short silent pause, : silent long pause, ‘+’: word cutoff, : sound lengthening (i.e. quattro)

Chapter Thirteen

252

SELF-REPAIR REPETITION Reparandum Editing Phase

Repair

Falling F0 Rising F0 -Prominence

Falling f0 Rising f0 Low intensity

Silent/Filled pauses Word fragments

Table 13-5: surface structure of Self-Repair Repetition. The pragmatic nature of this repetition is the self-repair of one or more lexical elements of the utterance: the lexical target is not temporally available, so the repetition highlights a problem with verbal planning. selfrepair presents, with similar percentages for R1 and R2, both falling and rising intonational patterns. The editing phase is always present and contains fragments of words, silent and filled pauses. Generally, the tonal excursion is low. Stalling repetition has a value of 29%. This pragmatic class displays only falling or level F0 contours. From an intonational point of view, both R1 and R2 are produced with a low intensity and a lowering of tonal values; the presence of a compressed pitch range is a peculiar feature of this repetition, as summarized in Table 13-6. STALLING REPETITION Reparandum Editing Phase Falling/level f0 Silent/Filled pauses - Prominence Compressed pitch range

Repair Falling/level f0 Low intensity

Table 13-6: Surface structure of Stalling Repetition. This is in line with the pragmatic function of the disfluency that is realized in order to take or fill time, see the examples (4-5). (4) e quindi lu+ lui/ lui è nervoso (5) forse tu sai / sì è quattro quattro è quattro giorni sì per quattro giorni The repetition stretched out over more time without conveying prominence or correction. The next two cases are quite interesting because R1 and R2 show a very different intonational strategy: in “expanding” type R1 has a prevalently low-rise F0 pitch (65%), while R2 is for the most part falling

Prosodic-Pragmatic Interface in L2 Italian Learners’ Repetitions

253

(80%); similarly, in emphasis type the most frequent F0 contour in R1 is low-rise (85%), but R2 has a high-fall F0 pattern (72%). Nevertheless, the two classes are not identical in either structure or function. The expanding repetition is a verbal mechanism employed to connect R1 with the rest of the utterance after a short fluency interruption (see Table 13-7). The learner hesitates and pauses, but is soon able to resume the utterance expanding its semantic topic, see example (6). (6) Cosenza Cosenza non mi piace perché sempre piove EXPANDING REPETITION Reparandum Editing Phase

Repair

Low-Rise f0 +Prominence Extended pitch range

Falling F0 - Prominence Compressed pitch range

Silent/Filled pauses

Table 13-7: Surface structure of Expanding Repetition. The last lexical item of R1 and the first item of R2 always coincide. For this reason, R2 is simply a continuation of the utterance. This is also proved by the presence of rising F0 contour in R1, that is the learner is realizing a typical continuation rise. When the repetition conveys prominence and emphasis (14% of all repetitions), R1 and R2 share some prosodic characteristics. In this case we found an extended pitch range (over 5 ST), high intensity and a very fast speech rate. A distinguishing feature of this repetition type is the systematic absence of the editing phase: R1 and R2 are contiguous, and both are prominent (see Table 138). Two other aspects corroborate the marked function of this repetition type: the longer duration of R1 and R2, with respect to other typologies already discussed, and the prevalence of perceptually salient F0 patterns. The learner is emphasizing the lexical content of the repeated words; as a consequence, all prosodic cues increase in order to achieve a suitable pragmatic prominence, see the examples (7-9). (7) Mi piace Italia molto molto (8) Perche io ho paura così così fish pesce piccolo piccolo pesci ma molto pesce (9) posso leggere bene bene bene

Chapter Thirteen

254

EMPHASIS REPETITION Reparandum Editing Phase Rising f0 empty High-Fall f0 + Prominence Extended Pitch Range High intensity Fast speech rate

Repair High-Fall f0 + Prominence

Table 13-8: Surface structure of Emphasis Repetition. In the last class, namely control over form (8%), we found only rising F0 patterns, but R1and R2 differed as regards to the other features. In particular, R1 is much longer than R2; it also presents a significant tonal excursion and high intensity values. On the contrary, R2 has a compressed pitch range and low intensity (see Table 13-9). The repeated word in R2 frequently has a low intelligibility, although the intonation pattern is preserved, F0 contour is similar in both R1 and R246, see the examples (10-11). (10) metto sempre nella piadella piadella? (11) non tutto pulito sì pulito? CONTROL OVER FORM REPETITION Reparandum Editing Phase Repair High-Rise F0 Rising F0 High-Fall F0 Extended Pitch range

Silent/filled pauses

Rising F0 High-Rise F0 Low intensity Compressed pitch range

Table 13-9: Surface structure of Control over form Repetition. This repetition shows a rising repair in line with its asking function. The learner in this way is asking for agreement. The repeated word in R2 is shorter than the same in R1, this confirmed previous findings:

46

The low intelligibility of the second occurrence of the same word has been already highlighted by Shields and Balota regardless of the repetition function. See Lynne W. Shields and David A. Balota, “Repetition and associative context effects in speech production”, Language and Speech 31 (1991): 47-55.

Prosodic-Pragmatic Interface in L2 Italian Learners’ Repetitions

255

“repetitions generally undergo phonological reduction processes which account for their lower intelligibility ratings”47.

5.2 The Anglo-American learner The corpus of our Anglo-American informant (henceforth, L.) exhibits different and numerous forms of repetition, most of which are selfrepetitions showing an ample range of different functions according to the frequence of occurrence. One way to look at the data is to observe what is repeated. Our informant shows a kind of variation in the repeated items. She mostly repeats exactly the source element except in a few cases (i.e. the fragment of talk that is repeated). This happens with self-correction when she realizes that she might be using the wrong form, or when she checks with the recipients if the word used is correct giving an alternative form: “ho ascoltato ? Ho sentito?” Another interesting aspect is the position of the repetition, i.e. whether the repeat is contiguous or noncontiguous. In L. corpus the source and the repeat are in some cases separated by a gap or some intervening talk,48 and not in immediate succession: “perché non penso troppo quando bevo vino non penso troppo e io parlo”, “molte famiglie che ha legami in Italia molte molte molte”. L. repetitions are mostly characterized by consecutive repeated items, without interference from the other party. The type of interference from the interlocutor is mainly aimed at correction or feedback over comprehension with a clarification check. Among the candidate functions of repeats in L. corpus (see Table 1310) we find: A. self-repetition a. Self-repair b. Control over comprehension and linguistic form control over linguistic form (check and signalling help over production from the recipient) c. Emphasis d. Stalling B. Hetero-repetition 47 Curl et al. “Repetition and the prosody-pragmatics interface”, 1721-51; Lynne W. Shields and David A. Balota, “Repetition and associative context effects in speech production”, 47-55. 48 Mirka Rauniomaa, “Recovery through repetition. Returning to prior talk and taking a stance in American-English and Finnish conversations”, academic diss., Faculty of Humanities of the University of Oulu (Oulu: Oulu University Press, 2008), 60.

256

Chapter Thirteen

Function Self-Repair Stalling/Practice Control over form or comprehension Emphasis Confirmation

% 22 25 12 24 17

Table 13-10: Percentage of repetition types. Among self-repetitions we can include the function the non-native speaker orientates towards her own production turn. It is called heterorestructuring,49 in which she produces an item or an expression and then abandons it giving the turn, causing a replay by the native who provides the correct form in the subsequent turn. Her answer (which has a didactic character) is then recovered by the learner with a hetero-replay, i.e. an exact repetition of the native to signal comprehension: (12) L: il mio paese lavoriamo troppo si ma in Italia la gente trova il tempo per risposo. M: Riposo? riposo L: riposo si si si, secondo me secondo me Sometimes the non-native simply continues, without the explicit help of the speaker, if she realizes that the interlocutor has got the point: (13) L: perché è una passione pass+ passione? anche ho studiato la lingua latina e mi piace molto studiare la lingua latina la letteratura latina, ma non parliamo la lingua latina so. (14) e quando ero bambina ehm sono andata nella cucina e guardarla e la guarda? (15) a miei prof esso+ resse, le mie prof alle mie professoresse 49 Jean François De Pietro, Marinette Matthey and Bernard Py, “Acquisition et contrat didactique: les séquences potentiellement acquisitionnelles dans la conversation exolingue”, in Actes du 3e colloque régional de linguistique, eds. Dominique Weil and Huguette Fugier (Strasbourg: Presses de l’Université de Sciences Humaines, 1989), 99-124.

Prosodic-Pragmatic Interface in L2 Italian Learners’ Repetitions

257

(16) So molti anni fa molti anni fa ha traduco tradutto (17) Parlo in italiano uhm quando frequente frequento le mie le mie (18) hanno trasl traslo they moved cambiano casa they moved into this house but it was difficult it was very very difficult for them. In the examples in (13) and (14), the production is orientated towards the correctness of her production and she repeats herself until she does not find the correct agreement. It is a conscious strategy to achieve proficiency, the correct linguistic expression, but at the same time she wants to know by her interlocutor if she has been understood: (19) ero bambino /ero bambina, ho ascoltato eh? Ho sentito? . As other investigations show, repetition orientated towards selfrepetitions of grammatical forms, which for the most part conscious tend to grow with the level of proficiency.50 Among our informants, L. seems to show more repetitions whose function is the control over production (meaning and form) than the group of informants with a lower level of proficiency. As Wiberg51 underlines, self-repetition in searching of the correct grammatical form, found before the turn is given, is a sign of selfreflection and access to the built-in grammar. Repeats are a way to learn something more from the conversational setting. Another function of the repeats present in the corpus is stalling or a sort of taking time in order for the speaker to organize her ideas, a function that allows time to the interlocutor. In (13), the second repetition lingua latina or simply latina serves to help the learner to continue participation in the conversation when she realizes that she has nothing really important to say.52 Most of the time, L. utilizes repetition as a form of iconism to express emphasis or intensity, often caused by a lack of linguistic competence of lexical materials. This type of repetition can also be described as metarepetition a sort of ratification of something she said, to show a repeated action. Very often, repeats (both self and other repeats) are used to double up the illocutionary force to emphasize or to persuade. The increase of

50

Wiberg, Il riferimento temporale nel dialogo, 318. Ibidem, 14. 52 Tannen, Talking voices. 51

258

Chapter Thirteen

form adds up to the increase of the illocutionary force,53 as we can see in the following examples: (20) L: molte famiglie che ha legami in Italia molte molte molte e c’è, c’è ci sono strade con on which abitano solo italiani solo italiani. (21) non è italiana, non è italiana, solo la famiglia del mio papà e la famiglia di mio papà è più aperta è più gentile è più rispettosa (22) L: bianca bianca bianchissima bello bello bello e i vicini would say would comment how white the sheets were. (23) L.: she would bake all the time, bread pane biscotti tutti i giorno tutti i giorno e quando ero bambina (24) L: Sì sì il la loro casa sempre sempre sempre è sempre oper aperta sempre any time of the day. The same function seems to hold in L1, as L. shows when she switches to her mother tongue: (25) L.: if they help you, then they will keep reminding you that they’ve helped you they’ve helped you they’ve helped you. In some way, the following type or repetition also seems to indicate a sort of diagrammatic relation between form and function, where the second item means a sort of routine action in which the intended meaning is making two things at a time: (26) pranziamo durante facciamo un giro a piedi pranziamo durante il lavo+ lavoriamo54. Confirmation-type markers are used to agree with the interactant, or to confirm what has been previously stated by the conversational partner. The prosodic investigation of the Anglo-American subject reveals significant changes according to the intonational and durational behaviours. As already discussed, in this subject the pragmatic repertoire of the functions conveyed by repetitions is wider. Despite her better proficiency of Italian language, this informant shows that typical foreign 53

Fuhui Hsieh, “Repetition in social interaction: a case study on Mandarin Conversations,” International journal on Asian language processing 19 (2009): 153-68. 54 The speaker means that Americans eat while they walk and during work hours, whereas Italians spend a lot of time just at lunch.

Prosodic-Pragmatic Interface in L2 Italian Learners’ Repetitions

259

accent that characterized all learners, even after long periods of exposure to a second language. Nevertheless, the management of prosody in this Anglo-American subject is more native-like with respect to that of A1 learners. As a consequence, the prosodic features are more differentiated and, in a certain sense, closer to those of the natives, although non-native speakers tend to resort to repetition strategy more frequently than native speakers.55 This is a general overview of the different communicative functions of repetition in the corpus of a non-native speaker of Italian used in conversation. It is important to consider the overall range of possibility to use repetition as a linguistic device to meet the communicative need in interaction. Disfluency of speaking skills, as we can notice from our corpora, is indeed a resource that language learners can utilize to enable them to engage in a conversation despite their language constraints. The functions here reported have to be analyzed with respect to the ones observed in the other corpus of learners.

5.3. Comparative analysis of learners An overall comparative analysis shows that some types of repetition do not equate exactly in the two corpora. As for the stalling type, it seems to entail different functions for the two level learners. For the A1 speakers it mainly indicates a strong difficulty in planning, and is characterized by a large number of pauses and word cut-offs. In our Anglo-American speaker, this strategy seems more native-like, and it is used either to organize discourse or to take time to recall a missing word or expression, to hold the floor while simultaneously thinking what to say next. Furthermore, the learner is engaged in a sort of training to practice a certain structure and therefore used as an overt control on her own production. Another sort of control which signals the need for cooperation by the recipient is control over form or comprehension. By means of repetition, the learner solicits a verification from her interlocutor. They are, therefore, an important tool used by the learner to ask for confirmation or clarification of meaning (see examples in (1), (2) and (3) or even in (8)). Through the use of repetition, the learner frames these events as both conversations and learning events.56 This kind of conscious learning, 55

Shields and Balota, “Repetition and associative context effects in speech production”, 47-55. 56 Silva and Santos, “Framing participation through repetition: the case of a Portuguese learner in different settings”, 17.

260

Chapter Thirteen

again, is not present at a lower level of proficiency, or at least it is not conveyed by means of repetition. Another difference among learners is the collocations of the repeated items. The Anglo-American production is more orientated through interaction and the response of the interlocutor than the production of the other learners. Repeats occur after other material intervening between the source and the second item, which can occur in order to redirect the attention of the interlocutor who did not attend the source utterance. Since we do not have sufficient examples of this type, we cannot formulate a clear function realized with these types of repeats. As for the type of item repeated, there is another difference between the two learners which is strictly related to their level of proficiency. Other than single lexical items, L. also produces clause repetitions, i.e. larger linguistic chunks within the same turn with or without the interruption of other lexical materials. On the other hand, Turkish and Spanish learners repeat mainly lexical items except for a few instances of collocation.

6. Conclusions Our analysis of repetitions has tried to shed some light on the relationship in the interface prosody/pragmatics as a means of investigating the intricate and subtle phenomenon of repetition. Our preliminary work has tried to establish some of the prosodic cues that helped us to individuate some of the parameters through which is possible to see how repetition is used and intended by non-native speakers of Italian. We hope to have shown that if repetition is to be understood as a resource in social interaction, it should be studied in connection with the practices and actions that it is used to implement. Such an understanding of language also entails an acceptance of the fact that interaction is inherently contingent. Participants may either modify their conversational moves or discard one and embark on another, in accordance with what others are doing.

PART IV TECHNOLOGY AND PROSODY RECOGNITION

CHAPTER FOURTEEN AUTOMATIC PROSODIC COMPARISON BETWEEN MODEL AND IMITATION SENTENCES IN A SECOND LANGUAGE TEACHING COMPUTERIZED ENVIRONMENT PHILIPPE MARTIN UNIVERSITY DIDEROT – PARIS 7, FRANCE

1. Introduction Sentence intonation, acoustically described by fundamental frequency, syllabic duration and intensity, is generally correlated with three linguistic objects: 1. Sentence modality (declarative, interrogative and their variants); 2. Information structure (highlighting important or new vs. old information); 3. Prosodic structure (as hierarchical organization of stress groups).1 1. Sentence modality is usually indicated prosodically by specific melodic contours located on the last stressed syllable or on the last syllable of the sentence. In Italian, declarative modality contours is located on the last stressed syllable, whereas interrogative contours are placed on the last syllable, whether stressed or not.

1

See Philippe Martin, “L’intonation de la phrase dans les langues romanes: l’exception du français”, Langue française mars 2004: 36-55; Philippe Martin, “Prominence detection without syllabic segmentation, Prosodic Prominence: Perceptual and Automatic Identification”, (paper presented at Speech Prosody, 5th International Conference - Satellite Workshop, Chicago, Illinois, May 11-14, 2010); Philippe Martin, Intonation du français (Paris: Armand Colin, 2009).

264

Chapter Fourteen

Example: Declarative: Marco è partito

Interrogative: Marco è partito

Many variants of basic modalities can be considered, together with their associated melodic contours.2 2. The information structure (IS) uses various prosodic mechanisms to either highlight specific stress groups (i.e. words), or mark specific syntagma as conveying new or old information. It can be instantiated with various acoustic features such as syllabic rhythm or emphatic stress located on the first syllable on non-grammatical words. Example: Mar-co è partito with a pause between the two syllables of Marco will put emphasis on Marco. Indication of segments carrying new vs. old information is realized with a so-called theme-rheme prosodic division of the sentence: Example: Marco è partito with a conclusive falling declarative melodic contour on the first and stressed syllable of Marco, whereas the following text segments carry flat contours on their stressed syllable: Mar\\ co è parti==to (\\ indicates a low and falling melodic contour, and == a flat melodic contour). 3. The prosodic structure (PS) reflects the strategy chosen by the speaker to help (or not) the listener to elaborate a hierarchical concatenation of the stress groups perceived in sequence (a stress group is defined by a sequence of syllables with only one stress). This hierarchy dynamically built helps the listener to elaborate a syntactic structure and allow him or her access to the sentence meaning though a complex process involving lexical access. The elaboration of the PS by the listener involves two processes: 2

Emanuela Cresti, Philippe Martin and Massimo Moneglia, “L’intonation des illocutions naturelles représentatives; analyse et validation perceptive”, in MacroSyntaxe et Pragmatique, L’analyse linguistique de l’oral, ed. by Mathieu Avanzi XXX (Roma: Bulzoni editore, 2003), 243-94.

Automatic Prosodic Comparison between Model and Imitation Sentences 265

a.

The conversion of sequences of syllables into higher rank linguistic units (usually a noun, verb, adverb or adjective, together with grammatical words such as conjunctions, pronouns, etc.), a conversion triggered by a stressed syllable in the sequence. In Italian, the position of stressed syllables also marks morphological boundaries. Example; ‘capitano vs. capi’tano in sono cose che capitano capitano (those are things that happen, captain) b. The hierarchical concatenation of stress groups elaborated in a) through decoding of specific melodic contours located on stressed syllables. These contours are differentiated and realized according to a grammar specific to each language.

Fig. 14-1: ITA Alcuni edifici si sono rivelati pericolosi.

Fig. 14-2: FRE Certains bâtiments se sont révélés dangereux.

Figures 14-1 and 14-2 show the melodic curves pertaining to two read sentences syntactically very similar in Italian and French. Although the

266

Chapter Fourteen

prosodic structures are identical in both examples, the sequence of melodic contours located on stressed syllables is not the same: the Italian prosodic structure assigns a high and rising melodic contour on the first stressed syllable (Alcuni) and a flat contour on the second stressed syllable followed by a rising contour on the last syllable (edifici), whereas the French prosodic structure first and second contours are respectively falling and rising (Certains bâtiments) while the third and last contours are very similar in both cases. In the process, the listener decodes stress groups not only from the nature of the melodic contour located on its stressed syllable, but also from rhythmic properties of syllable sequences, and from known syllabic pattern stores in his or her memory. Rhythmic patterns, involving the alternation of short and long syllables, the stressed syllable being generally the longest, are particularly important in identifying syllabic groups already known to the listener. Together with melodic curves located on stressed syllables, it is therefore equally important for the acquisition of Italian sentence intonation by L2 learners. The linguistic intonation briefly examined above clearly indicates the importance for an L2 learner to correctly master the prosodic system of the language involved. Actually, it has been shown3 that the prosodic structure operates the last restructuring resulting from three preceding successive processes: phonological, morphological and syntactic. Conversely, the prosodic structure is the first hierarchical information used by the listener before syntactic decoding. Teaching L2 oral competence therefore initially implies a reasonable mastering of the prosodic properties of the language for better comprehension by L2 listeners.

2. Computer aided language learning (CALL) Using computer-aided oral teaching is by no means a new approach. An early, if not the earliest, implementation of a device graphically showing intonation parameters such as the melodic curves of both a teacher and the learner imitation has been described in Vardanian.4 With the advances of computer technology, other realizations appeared.5 3

Claire Blanche-Benveniste, “La naissance des syntagmes dans les hésitations et répétitions du parler”, in Le sens et la mesure. Hommages à Benoît de Cornulier, ed. Jean Louis Araoui (Paris: Honoré Champion, 2003), 40-55. 4 Rose Marie Vardanian, “Teaching English through oscilloscope displays”, Language Learning 3/4 (1964): 109-11. 5 See among others Harlan Lane and Roger Buitenton, “A Self-Instructional Device For Conditioning Accurate Prosody: Trends”, in Language Teaching, ed.

Automatic Prosodic Comparison between Model and Imitation Sentences 267

Although implying sophisticated hardware and software schemes, these realizations lack phonological insight in their implementations. Indeed, the common and classical layout of these devices involves some graphical display of a teacher and learner melodic curves, sometimes together with other acoustic information such as speech intensity and wave form (see Fig. 14-1 for a commercial implementation).

Fig. 14-3: The PM100 displaying model and imitation melodic curves (Voice Identification, 1982).

The main underlying idea of these realizations is that learners are supposed to be “deaf” to the phonological system to be acquired (as their perception is mainly driven by their mother tongue phonological grid), and that their pronunciation could be tremendously improved if the prosody of their realization was approaching the target language prosody. If learners were considered “deaf” to certain phonological or phonetic features, giving them supplementary graphical information was felt, at the time, to be highly beneficial. These assumptions reveal the paramount importance given to sentence prosody in the pronunciation of a foreign language, as well as the benefit of graphical input to supplement auditory perception. Real time visual feedback pertaining to the three main parameters of prosody, fundamental frequency, intensity and syllable duration, became an essential part of CALL development to acquire the prosodic features of a learned language. Many hardware and software developments implementing these ideas appeared in the last 40 years, along with the progress made in computer A. Valdman (New York: Academic Press, 1969), 159-74; Eric James, “The Acquisition of Prosodic Features of Speech Using a Speech Visualizer”, IRAL 14 (1976): 227-44; Dorothy M. Chun, “Signal analysis software for teaching discourse intonation”, Language Learning and Technology 2/1 (1998): 61-77.

268

Chapter Fourteen

technology. Among the most notable, were Madsen (1973), VisiPitch (1975), Pitch Computer (1978), Speech Viewer (1985), for hardware devices; and WinPitch (1996), Speech Tutor (2003) for software packages. All these proposed, to the learner, an imitation of some model intonation curve, without reference to any phonological or phonetic fact (with the possible exception of sentence modality supposedly encoded by the sentence final intonation contour). This drill method approach6 does not pertain to any linguistic knowledge of intonation, which may explain why the use of pitch visualizers did not become very popular during three decades of development, as their effectiveness was debatable.7 Another aspect that hampered a large use of these devices is the apparent complexity of the so-called pitch curve, an acoustical estimation of the time evolution of the laryngeal frequency. Not only have these curves reflected a particular phonetic realization of some intonation model, but graphic details displayed on screen are often not pertinent linguistically or simply not perceived by the listeners. Better ways of displaying the prosodic information have been sought,8 but only in integrating perception effects in displaying the prosodic information. Another problem in the development of teaching applications is linked to the graphic emphasis on pitch at the expense of rhythm, which is another important component of prosody. Speech synthesis experiments have shown that in some cases rhythm is more important for comprehension than pitch. An example of application that puts emphasis on rhythm rather than pitch can be found in Delmonte.9 The problem with such a layout is the lack of linguistic information given to the learner while attempting to imitate a rather complex melodic curve. No provision is made for highlighting the important segments of prosody to be correctly imitated, and the layout does not make any provision for any melodic variants that could be considered acceptable. The learner is given a task rather like acquiring the complete morphological system of L2 by merely listening to a large number of examples but without receiving any grammatical information. Although this process might work, given long enough, it is of course much more 6

Joseph D. O'Connor and F. Arnold Gordon, Intonation of Colloquial English (London: Longman, 1973). 7 Kees de Bot, “Visual feedback of intonation I: Effectiveness and induced practice behaviour”, Language and Speech 26/4 (1983): 331-50. 8 See Gerard Spaai and Dik Hermes, “A visual display for the teaching of intonation”, CALICO Journal 10 (1993): 19-30. 9 Rodolfo Delmonte, “Prosodic tools for language learning”, International Journal of Speech Technology 4 (2009): 161-84.

Automatic Prosodic Comparison between Model and Imitation Sentences 269

time efficient to provide relevant grammatical information. The same conclusion is valid for sentence intonation. Without any explanation as to which are the important time segments (syllables) to be correctly imitated and why they are important, and which are the less important ones, WHO will give the learner a better chance to achieve his or her goal in mastering the L2 sentence intonation, essential in ensuring correct and easy comprehension by native speakers, an improvement in learner performance which is definitely not cosmetic.

3. Rationale for CALL The development of WinPitch LTL as a tool to teach the prosody (and possibly other phonetic features) of a foreign language was conducted along the following well known pedagogical line going from passive to active learning: 1. I hear and I forget 2. I see and I remember 3. I do and I understand.

3.1. I hear and I forget “I hear and I forget” corresponds to the situation found in early language laboratories: the learner simply repeats the model heard, and sentences can be organized in sequences reflecting the acquisition of a particular point of pronunciation (e.g. the mute e or vowel-vowel liking in French). Prosodic aspects were often limited to the location of lexical stress and the correct pronunciation of stress groups. Only in tone languages such as Mandarin would the realizations of pitch acquire some phonological significance in the drills offered to the learner.

3.2. I see and I remember “I see and I remember” is reflected by the advent of pitch visualizers, displaying, sometimes in real time, a pitch curve model to be imitated by the learner. The advantage of a simple “listen and repeat” approach pertains to the assumed phonological deafness of learners, who can now see what they might not hear. The possibility of slowing the speech rate of the model down constitutes a further improvement, allowing the correlation between auditory and visual perception: the learner can now link perceptively the graphic movement of pitch on the screen with the

270

Chapter Fourteen

perception of it (through the synchronous displacement of a screen cursor, for instance).

3.3. I do and I understand Through the modification of the prosodic parameters (fundamental frequency, intensity, syllable duration, rhythm and pauses) of their own voice with re-synthesis techniques such as PSOLA, the learner is now able to graphically manipulate his or her production and therefore get a direct understanding of how to achieve the prosodic movements required on their own voice, without actually performing it themselves. This corresponds to a “learning by doing” learning process.

3.4. I Automated feedbacks The use of pitch visualizers can be in praesentia, with the possible comments of an instructor who can directly intervene to guide the learner in the process, or in absentia, with the software automatically making its own comments on a particular learner performance. In commercial realization, this feedback is more than often limited to a score, with no or very little comment of the type and place of errors made in the learner imitation. Obviously this feedback should reflect specific points of an (underlying) intonation theory.

3.5. Prosodic morphing In an L2 sentence intonation learning environment, prosodic morphing operates by transferring melodic movements, (inter)syllabic durations, and possibly intensity values from a model sentence (usually part of a set) to a learner imitation sentence. As sentence overall duration and (inter)syllabic durations usually do not correspond, fundamental frequency and durations values are aligned either on syllable boundaries or on syllabic energy peaks. If properly processed, i.e. without perceptively notable syllabic segmentation or fundamental frequency errors, the learner realization will be played back with most prosodic features of the model, therefore with the learner’s own voice but with a corrected sentence intonation. Prosodic morphing may then not only convince the learner that a correct intonation can indeed be obtained, but, what is more important, where the prosodic correction should be made. Indeed, in the prosodic morphing implemented in WinPitch, the learner can visualize on his or her own voice acoustic analysis the differences in syllabic and intersyllabic duration (responsible

Automatic Prosodic Comparison between Model and Imitation Sentences 271

for rhythm) as well as melodic curve. If a sound grammatical explanation is given at the same time, i.e. the reason why a particular melodic curve rises on a specific syllable or why a specific syllable is shorter or longer in a certain context, the effectiveness of the method for L2 sentence prosody acquisition could be extremely beneficial. What is missing in most, if not all, these realizations is a clear reference to any linguistic content (phonological and/or phonetic). Learners were put in front of a computer screen displaying information that was hard or impossible to link to any coherent intonation theory. The acquisition of Mandarin tones is a good counter example where the role of pitch is clearly phonological. Therefore its description and integration in a systematic set of examples can be more easily established. The automatic feedback given to the learner can therefore be phonological as pertaining to the system of contrasts existing between the four tones of the language, and phonetic in commenting the particular details of realization in term of syllable length and pitch contour.

4. Implementation The current version of WinPitch LTL attempts to implement these principles: it gives the learner the opportunity to listen to models at a reduced, programmable speech rate to enhance comprehension, to see the prosodic parameters on screen either in real time or at leisure to listen to his/her performance and to learn by doing by modifying the prosodic parameters of his/her own voice graphically according to the prosodic features to be acquired. Phonologically pertinent speech segments are displayed, highlighted in a colour chosen by the author-teacher responsible for putting model sentences together. Furthermore, the automated prosodic morphing allows the learner to listen to her or his own voice and understand why his or her modified performance is enhanced, by comparing visually on screen the prosodic modifications made for the fundamental frequency and duration on each syllable.

4.1. Navigation WinPitch navigation toolbar allows for an easy selection of sentences in a lesson, to listen to the model at a variable speech rate (programmable in the 15% - 200 % range), to replay learner imitations in any order, to listen and repeat in sequence all the models of a lesson, and to graphically enter prosodic modifications of any learner repetition.

272

Chapter Fourteen

Fig. 14-4: WinPitch LTL toolbar.

A zoom function displays the model on the whole screen with or without the corresponding spectrographic display of both model and sentence. The screen is divided in a top section reserved for the model, the aligned text, the model highlighted pitch and intensity curves (a spectrogram can also be displayed), and a bottom section allocated for the learner imitation.

Fig. 14-5: Model spectrographic display.

4.2. Slow rate playback To ensure a better perception of the model by the learner and allow the visual and auditory coordination between a screen cursor and speech played back at a slow rate, the model and imitation playback rate can be adjusted continuously between 15% - 200%. Slow playback is performed by a PSOLA10 engine, based on the pitch synchronous insertion or deletion of pitch period. Thanks to the use of a reliable pitch tracking algorithm (the spectral comb method), the slowed speech is of excellent quality. Conversely, the model can be played back at a higher speed to test learner comprehension in various conditions.

10

Eric Moulines and Michel Charpentier, “Pitch-synchronous waveform processing techniques for text-to-speech synthesis using diphones”, Speech Communication 9 (1990): 453-67.

Automatic Prosodic Comparison between Model and Imitation Sentences 273

4.3. Speech segment highlighting Speech segments can be highlighted by the author in any colour, allowing him to define the linguistically or otherwise pertinent sections of the model to be imitated by the learner.

Fig. 14-6: Model stressed syllables highlighting, the pertinent stressed segments are highlighted (circled in this example).

In Fig. 14-6, stressed syllables of the sentence “When Pierre was in London” are highlighted in red in the software and circled in the figure, showing the contrast of pitch slope rising falling typical in English in sentences with two stress groups.

4.4. Prosodic morphing Figures 14-7 and 14-8 show a simple example with successive prosodic morphing operations: syllabic segmentation, (inter)syllabic staircase duration curve, (inter)syllabic duration morphing, fundamental frequency (F0) display and morphing. Syllable peaks are indicated by vertical lines, and duration by horizontal lines. One can easily notice the differences in rhythm and melodic curve. The first stressed syllable on nessuno has a high and rising melodic contour, whereas the learner realization shoes a middle and falling contour. All four prosodic parameters (fundamental frequency, intensity, syllable duration and pauses) can be modified on either the model or the learner sentence through simple graphic commands. Using the spectrographic display, the user can easily change specific syllable durations or pitch movement according to the model presented to the left of the screen. Propositioned duration (in red in the software, in grey in the fig. 14-9) and pitch (in white) lines are placed on the learner part of the screen. These lines can be dragged, cut, and its vertices easily placed with the mouse on an appropriate spot on the screen in order to define the new duration and

274

Chapter Fourteen

pitch pattern obtained during re-synthesis (activated by clicking on a single button).

Fig. 14-7: A simple example: nessune di queste ragione riguardava la moglie. Top: original model; bottom: learner imit3ation, French speaker.

Fig. 14-8: The same example displaying the simplified melodic curves for both realizations.

Automatic Prosodic Comparison between Model and Imitation Sentences 275

Fig. 14-9: F0 and duration morphing, both duration and fundamental frequency values are imported from the model, and positioned between corresponding syllable peaks.

There is no guarantee that the learner will pronounce the same number of syllables present in the model, or that the learner syllabic realizations will result in the same segmentation, corresponding to the model segmentation, with the same number of syllables for both model and imitation. In WinPitch implementation there is therefore a warning message displayed automatically when the numbers of syllables of the model and the imitation do not match.

4.5. Authoring mode An authoring mode of the programme gives total control of all functions to elaborate sets of model sentences. From a pre-recorded file (in wav or mp3 format), or after a direct recording of model sentences, the author can, with simple commands, insert the segments reserved after each model for the learner production, add text on the notepad and on screen, pre-place the duration and pitch morphing lines, highlight pertinent segments with any color, define segments for automatic mapping on the learner imitation, and add comment in HTML format, etc.

276

Chapter Fourteen

5. Conclusions WinPitch LTL is an innovative software programme derived from older versions developed in the last 12 years.11 To the learner, it offers a user-friendly interface, allowing easy navigation in the sets of model sentences included in a lesson. The learner can listen to the models proposed at variable speech rate to obtain a convincing correlation between visual and auditory perception of pitch movements and syllable durations related to pertinent segment highlighted in colour, and repeat and observe his or her own realization in terms of prosodic parameters displayed in real time. Model pertinent highlighted segments are automatically mapped on the imitation, and the learner can, if necessary, correct the corresponding pitch, intensity and duration parameters through simple graphic command driving re-synthesis of his/her own voice. Thanks to the use of a complete set of authoring functions, the teacher can easily build lessons by defining graphically time spaces reserved to the learner, by adding text in any language (using Unicode fonts) on screen and use the notepad to write notes in HTML format. The software can be downloaded from http://www.winpitch.com.

11

Aline Germain and Philippe Martin, “Présentation d’un logiciel de visualisation pour l’apprentissage de l’oral en langue seconde”, Apprentissage des Langues et Systèmes d’Information et de Communication 3/1 (2000): 61-76.

CHAPTER FIFTEEN TEXT-TO-SPEECH SYNTHESIS: PHONETIC MAPPING APPLIED TO ITALIAN AS TARGET LANGUAGE PIER LUIGI SALZA, ALBERT DE GROOT AND ENRICO ZOVATO LOQUENDO S.P.A. – TORINO, ITALY

1. Introduction 1.1. Concatenative TTS synthesis Loquendo Text-To-Speech (LTTS®)1 is a multilingual, multi-voice TTS system based on the concatenative synthesis technique, relying on contextual variable-length unit selection. The basic idea of unit selection is an evolution of synthesis by diphones. For each voice, units are not physically defined once and for all, but rather they are defined and selected at run time from a large speech database, recorded by a professional speaker in his native language. The richer the speech database, the better the synthesis quality. Every phonetic sequence in every prosodic context, for the given language, is collected in the database. The aim of this technique is to recombine natural voice fragments by retaining their original acoustic, phonetic and prosodic content, so as to avoid any further speech signal modification. The design of the database to be recorded relies on accurate (written) language domain statistics, combined with phonetically dense text selection done by means of “greedy algorithms”, in order to obtain the optimal language domain coverage. The 1

Marcello Balestri et al., “Choose the best to modify the least: a new generation concatenative synthesis system” (paper presented at EUROSPEECH ’99, Budapest, Hungary, September 5-9, 1999).

278

Chapter Fifteen

recorded speech database is then digitally converted and automatically segmented and analyzed. At the end of this process, each phoneme is identified by phonetic, acoustic (F0, duration and intensity values) and prosodic labels. The speech database represents a collection of potential elements (every phoneme of a given language/voice available in every phonetic context and in different prosodic variations) to be defined at run time, at each selection needed to generate a given sentence. The selection is carried out according to the following criteria: x Phonetic correspondence between the input string and the string found in the database. x Length of the phonetic sequence. x Prosodic label identity. x Preferred cutting points on consonants. The main advantages of contextual variable-length speech units over diphones are the following: x Longer units reduce the number of junctions and thus the possible occurrence of discontinuities, enhancing intelligibility. x The availability of as many different occurrences of each phoneme as the different prosodic contexts, instead of a single sample, avoids further signal processing, enhancing naturalness. x Simple waveform concatenation of specific units close to the prosodic target preserves the original shape of the intonation curve, avoiding distortions; for smoothing any residual abrupt F0 changes at unit junctions, a percentage scaling algorithm acts between contiguous units. The working flow of LTTS® is sketched in Fig. 15-1. Roughly, the text-to-speech process involves two main steps: 1) Text translation into structured linguistic information, including eventual Automatic Phonetic Mapping and including prosodic boundary marking; 2) Signal generation, including selection and concatenation of the more suitable acoustic segments for generating speech. Automatic Phonetic Mapping (APM), illustrated with more details in section 2, enables each voice (L1) to pronounce every recognized-as-

Text-to-Speech Synthesis

279

foreign word or sentence (L2) by using its own phoneme set. Every voice is able to speak every language supported by the system.

Fig. 15-1: Block diagram of LTTS®.

1.2. Phonetic labelling and grapheme-to-phoneme transcription Accurate phonetic labelling and transcription represent the key requirement for the effective operation of the APM system, automatic phonetic segmentation and external lexicon access as well. LTTS® labelling is based on phonetically justified choices and has a high degree of accuracy, allowing, at the same time, the unambiguous representation of every phoneme of the various languages (language independent) and of every allophonic variant of each single voice (speaker dependent).2 Unambiguous transcription code is guaranteed by the association of each phonetic symbol to a string of articulatory-phonetic features and to a prosodic label.

2

Pier Luigi Salza, Claudia Barolo and Beata Dobrzynska, “Etichettatura fonetica e TTS” (paper presented at the IV Congresso Nazionale AISV, Arcavacata di Rende, Italy, December 3-5, 2007).

280

Chapter Fifteen

Phonetic symbols are written in XSAMPA3 format. Articulatoryphonetic features derive from IPA (International Phonetic Alphabet). They allow the following classifications: x VOWELS: non-binary categories (position of the articulator on horizontal and vertical axes) and binary categories (nasalization, rotacization, stressing, and so on). x CONSONANTS: non-binary categories (manner and place of articulation) and binary categories (voicing, aspiration, release, syllabicity, and so on). Table 15-1 shows some examples of LTTS® phonetic labelling.

Table 15-1: Examples of LTTS® phonetic labelling, in IPA and XSAMPA symbols. 2: second vocalic segment in a diphthong. Prosodic labelling is related to the phoneme position with respect to prosodic boundaries. For carrying out grapheme-to-phoneme transcriptions, two methods are used: Lexicons and Contextual Rules. Lexicons (a list of words with the corresponding phonetic transcriptions) produce accurate transcriptions, although they require massive memory resources. Moreover, they will not be exhaustive. They thus need to be combined with statistical approaches in order to predict the transcription of words not present in the lexicons. Contextual Rules, capable of predicting the correct pronunciation of words, do not require large memory resources, but can nonetheless reach 3

John C. Wells, “SAMPA computer readable phonetic alphabet”, last revised October 25, 2005, http://www.phon.ucl.ac.uk/home/sampa.

Text-to-Speech Synthesis

281

high levels of complexity, though they require the management of pronunciation exceptions. Owing to the drawbacks of the single adoption of one or another of these two approaches, the optimal solution often relies on a combination of the two methods, with various degrees of prevalence according to the language. For instance, in the case of phonologically regular languages, like Italian or Spanish, the conversion is mainly done by rules; while for English, a phonologically irregular language, automatic transcription is mainly based on lexicons.

2. Automatic Phonetic Mapping Although text-to-speech conversion is intrinsically language dependent, real applications of speech synthesis more and more often include mixedlanguage texts, namely texts where single words or entire sentences are loaned from different (foreign) languages. This is very common in English. In these cases, pronunciation must be able to adapt run-time to the foreign (L2) language, if recognized, while maintaining the same voice (L1). The LTTS® multilingual architecture, the accurate phonetic labelling and the capability of detecting foreign language words or sentences by means of a language guesser or specific markers inserted in the text, have all allowed the implementation of an Automatic Phonetic Mapping (APM) module.4 APM enables each voice (L1) of LTTS® to speak every language (L2) supported by the system, by using its own phoneme set. The aim is the reading of L2 words or sentences inserted in an L1 text with an approximate but plausible pronunciation (not exactly that of L2). In APM, the L2 phonemes missing in L1 speech database are mapped onto similar phonemes present in L1. In order to be able to run in every L1 to L2 combination, APM is based on a very abstract principle: two phonemes can be considered similar if they are defined by similar articulatoryphonetic features. Candidate phonemes are selected according to a phonetic distance function based on weightings assigned to each articulatory-phonetic feature. Defining the degree of similarity between the values of non-binary features (consonant manner and place of articulation, and vowel position on horizontal and vertical axes), has been a particularly challenging task. Non-binary feature values have been 4

Leonardo Badino, Claudia Barolo and Silvia Quazza, “A general approach to TTS reading of mixed-language texts” (paper presented at ICSLP 2004, Jeju, Korea, October 4-8, 2004).

282

Chapter Fifteen

arranged on a perceptual distance scale. Although APM has the advantage of being “universal”, in several cases the acoustic and prosodic quality of the concatenations generated by APM is a little poor. In Table 15-2, an example of L1 Italian to L2 British English APM mapping is shown, for the sentence: “Her party was a success”.

L2 text L2 XSAMPA L1 to L2 APM XSAMPA

L1 Italian, L2 British English Her party was a success. h3:#p_h"A:4_di#w@z#@#sVk_}s"Es| e#p"ati#wez#e#seks"Es|

Table 15-2: Example of APM Italian female voice (L1) phonetic mapping onto British English (L2). Comparison between L2 standard transcription and APM transcription (in XSAMPA) of the sentence: “Her party was a success”; bold: mapped phonemes; bold underlined: phoneme fault; #: word boundary; |: prosodic boundary. The analysis of the speech sample shows a number of faults. The aspirated fricative [h], not present within the Italian phoneme set, has been deleted; the open-mid central vowel [3:] has been mapped onto [e], as has the schwa [@]; the open back vowel [A:] has been mapped onto [a]; the open-mid back vowel [V] again onto [e]. The main sources of unnaturalness are perceived in the [e] in place of [3:] and in the [a] in place of [A:]. In both these contexts, Italian listeners perceive as quite unnatural the lack of an alveolar trill. As the paper focuses on Italian as L2, in Table 15-3 an example of L1 French to L2 Italian APM mapping is shown for the sentence “Per parlare una lingua straniera, bisogna imparare le parole di questa lingua” (In order to speak a foreign language, it is important to learn the words).

Text-to-Speech Synthesis

283

L1 French, L2 Italian L2 text of P1 L2 XSAMPA of P1 L1 to L2 APM XSAMPA of P1 L2 text of P2

Per parlare una lingua straniera,

L2 XSAMPA of P2 L1 to L2 APM XSAMPA of P2

biz"OGna#impar"are#le#par"Ole#di#kw"esta#l"i Ngwa| biz"oGna#impaR"aRe#le#paR"Ole#di#kw"esta #l"iNgua|

per#parl"are#una#l"iNgwa#stranj"Era| peR#paRl"aRe#una#l"iNgua#stRanj"ERa| bisogna imparare le parole di questa lingua.

Table 15-3: Example of APM French male voice (L1) phonetic mapping onto Italian (L2). Comparison between L2 standard transcription and APM transcription (in XSAMPA) of the sentence: “Per parlare una lingua straniera, bisogna imparare le parole di questa lingua” (In order to speak a foreign language, it is important to learn the words). P1: phrase 1; P2: phrase 2; bold: mapped phonemes; #: word boundary; |: prosodic boundary. The phonetic output of APM is quite close to the original transcription, including the stressed/unstressed feature. The only replacements are: [R] instead of [r] and [u] instead of [w] in P1, and [o] in place of [O] in P2. Most phonemes useful for the “correct” transcription of the L2 sentence have been found in the L1 voice database. Nonetheless, the acoustic quality of the utterance is a little poor, as some acoustic discontinuities and unnaturalness are perceived. Yet they mainly affect the sequences containing [Gn] and [N], as shown in Table 15-4, where APM unit selection for the same sentence seen above is presented.

Chapter Fifteen

284

L1 French, L2 Italian L2 text of P1

Per parlare una lingua straniera,

L1 to L2 APM XSAMPA of P1

peR#paRl"aRe#una#l"iNgua#stRanj"ERa|

L1 to L2 APM units for P1

(| p) (p e R) (R p a R l) (l “a R) (R e) (e u n) (n a l) (l “i) (“i N g) (g u) (u a) (a s t R) (R a n) (n j) (j “E R) (R a)

L2 text of P2

bisogna imparare le parole di questa lingua.

L1 to L2 APM XSAMPA of P2

biz"oGna#impaR"aRe#le#paR"Ole#di#kw"esta# l"iNgua|

L1 to L2 APM units for P2

a | b i z) (z “o) (“o Gn a) (a i) (i m) (m p) (p a R) (R “a R) (R e l) (l e p) (p a R “O l) (l e d) (d i k) (k w) (w “e s) (s t a l) (l “i) (“i N) (N g) (g u) (u a) (a |)

Table 15-4: Units selected using APM for synthesizing the sentence “Per parlare una lingua straniera, bisogna imparare le parole di questa lingua” (In order to speak a foreign language, it is important to learn the words). L1 French male voice; L2 Italian language; P1: phrase 1; P2: phrase 2; bold: mapped phonemes; (----): unit; #: word boundary; |: prosodic boundary; bold underlined: APM units with acoustic and prosodic faults. Analysis of these faults shows that one of the reasons for low acoustic quality has to do with the low frequency of occurrences (in the L1 voice database) of the (units containing the) mapped phonemes. By looking at the perceived faults, occurrences are 122 for [Gn] and, respectively, 17 for [N]. These data are (also) explicable by accounting for the specificity of French (L1), a language where words are typically stressed on the last syllable. Thus, both stressed syllables not in end-of-word positions and unstressed syllable in end-of-word positions are likely to be few or rare. Systematic fault analysis of APM has been carried out on the following mappings: Italian to French, Italian to English, French to Italian and

Text-to-Speech Synthesis

285

German to Italian. The investigation highlighted that the faults are mainly due to some critical aspects of APM: x The rigid definition of the weightings given to each articulatoryphonetic feature will not be suitable for every language. As assessed by Nozawa and Cheon,5 “the perceived similarity of interlingual vowel pairs is affected by listeners’ L1 phonology as well as phonetic distance between the two vowels”. x Different articulatory-phonetic features may affect in different ways the “perception of similarity”: for example, heuristic experience seems to suggest that, in comparing vowels, variations of the rounding features (rounding degree) are more discriminating than variations in the lexical stress features (stressing degree: primary stress, secondary stress, unstressed). x The phonetic distance evaluation is strictly limited, in APM, to the phoneme-by-phoneme criterion and does not account for the frequency of occurrence of the candidate phoneme within the L1 voice acoustic database. This does not always allow to achieve the best selection, in terms of both acoustic unit extension (in number of contiguous phonemes) and acoustic parameter matching between units. Choosing from only a few candidates will probably result in a bad match at the boundaries, at least as far as F0 and intensity are concerned. The previous analysis suggested some ideas for improving the acoustic/prosodic quality of speech generated when automatic mapping is activated.

3. Experimental Phonetic Mapping According to the above considerations, APM output phonemes should be replaced when the following cases occur: x The phoneme of L2 is present in the speech database of L1 voice, but there are few occurrences of it. Clearly, it does not belong to the phonological system of L1 and its occurrences are due to loan 5

Takeshi Nozawa and Sang Yee Cheon, “How listeners’ L1 affects perceived similarity of American English, Japanese and Korean vowels” (paper presented at ICPhS XVII, Hong Kong, August 17-21, 2011).

286

Chapter Fifteen

words that the speaker had tried to pronounce according to L2 rules. x The phoneme of L2 is not present in the speech database of L1 voice, and the APM default mapping does not work well. Two heuristic principles have been defined for managing the substitution (repair rules): x In both cases, a “similar” phoneme, chosen according to specific phonetic knowledge of both L1 and L2 languages, if well represented in the L1 database, will sound better. This choice is consistent with the normal behaviour of the native speaker, who tries to pronounce the inserts of L2 while maintaining as much as possible his own phonological system. He instinctively operates the simplest substitution induced by articulatory characteristics of L1. In this respect, several studies address the notion of phonetic similarity6 or describe the variation of discrimination between nonnative phonemes, depending on how well each of them fits into native categories.7 Within these repair rules, an important role is played by the removal of the constraint related to the stressed/unstressed feature in the case of vowels, when it needs. It has been observed that a smooth concatenation with a “wrong” unstressed vowel sounds noticeably better than a rough concatenation with the “right” stressed vowel, and vice versa. Intelligibility has to be preferred to abstract correctness. x In both cases, the similarity has to be evaluated within a larger phonetic context, by removing the constraint of the phoneme-byphoneme criterion, in order to favour the selection of longer units. Although there are obvious drawbacks to this procedure, since the repair rules have to be defined language by language, its effectiveness has 6 Patricia K. Kuhl, “Human adults and human infants show a ‘perceptual magnet effect’ for the prototypes of speech categories, monkeys do not”, Perception & Psychophysics 50 (1991): 93-107; Jim E. Flege, “Assessing constraints on secondlanguage segmental production and perception”, in Phonetics and Phonology in Language Comprehension and Production: Differences and Similarities, eds. Niels O. Schiller and Antje S. Meyer (Berlin: Mouton de Gruyter, 2003), 319-55. 7 Catherine T. Best, Gerald W. McRoberts and Elizabeth Goodell, “Discrimination of non-native consonant contrasts varying in perceptual assimilation to the listener’s native phonological system”, Journal of the Acoustical Society of America 109 (2001): 775-94.

Text-to-Speech Synthesis

287

been checked by means of an Experimental Phonetic Mapping (EPM) prototype, developed for working off-line, for the following mappings: Italian to French, Italian to English, French to Italian and German to Italian. For this experiment, a total of 17 sentences and short paragraphs, representative of the several types of faults detected in APM, were selected among those analyzed during the investigation, and these have been generated by EPM as well. Sentences are reported in the Appendix. Then, the speech signals have been analyzed and compared with the corresponding APM versions. In Table 15-5, a comparison between APM and EPM transcriptions of the same sentence of Table 15-3 and Table 15-4 is shown. In end-of-word position, the French-style stressed syllables have been preferred to the unstressed ones. The analysis of the word “bisogna”, trascribed respectively [biz"oGna] in APM and [bizonj"a] in EPM, shows that: [Gn], as previously seen, has 122 occurrences while [n] has 2050 occurrences and [j] 1522. In the case of “lingua”, transcribed respectively [l"iNgua] in APM and [lingw"a] in EPM, the analysis of phoneme occurrences within the L1 database shows that: [N] has globally only 17 occurrences, as previously seen, a number which decreases a lot if stressed syllables are selected, while [n], as previously seen, has 2050 occurrences. In Table 15-6, a comparison between APM and EPM unit selection for the same sentence shown above is presented. Improvements have been obtained by selecting slightly different phoneme sequences, i.e. [n j “a] instead of [Gn a] and [i n g w ”a] instead of [“i N g u a], which are better represented in the L1 voice acoustic database. Although the words are last-syllable stressed, and despite [n] being chosen instead of [Gn], the perceptual effect of EPM is better. Speech sounds more fluid and the prosody (although that of L1) is more consistent. In addition to L1’s phoneme set, EPM preserves, at the suprasegmental level, L1’s prosody as well. This result is consistent with what occurs in real speech: an L1 speaker seems to pronounce L2 “insertions” by maintaining both his own phonological system and his own prosody.

288

Chapter Fifteen

L1 French, L2 Italian L2 XSAMPA of P1 L1 to L2 APM XSAMPA of P1 L1 to L2 EPM XSAMPA of P1 L2 XSAMPA of P2 L1 to L2 APM XSAMPA of P2 L1 to L2 EPM XSAMPA of P2

per#parl"are#una#l"iNgwa#stranj"Era| peR#paRl"aRe#una#l"iNgua#stRanj"ERa| peR#paRlaR"E#wn"a#lingw"a#stRanjeR"a| biz"OGna#impar"are#le#par"Ole#di#kw"esta#l"i Ngwa| biz"oGna#impaR"aRe#le#paR"Ole#di#kw"esta# l"iNgua| bizonj"a#impaRaR"E#l"E#paROl"E#d"i#kwe st"a#lingw"a|

Table 15-5: Comparison between standard L2 transcription and, respectively, APM and EPM transcriptions of the sentence: “Per parlare una lingua straniera, bisogna imparare le parole di questa lingua” (In order to speak a foreign language, it is important to learn the words). L1 French male voice; L2 Italian language; P1: phrase 1; P2: phrase 2; bold: replaced phonemes; #: word boundary; |: prosodic boundary. In this regard, the study by Boula de Mareüil and Vieru-Dimulescu,8 expands our understanding of what is meant by “foreign accent”, suggesting how prosody is important in identifying “foreign-accented” speech. The combination of fundamental frequency changes and the pronunciation of certain phonemes result in the perception of a foreign accent, while each cue taken separately may not. So, a foreign accent should reflect both segmental and suprasegmental dimensions.

8

Philippe Boula de Mareüil and Bianca Vieru-Dimulescu, “The contribution of prosody to the perception of foreign accent”, Phonetica 63 (2006): 247–67.

Text-to-Speech Synthesis

289

L1 French, L2 Italian L1 to L2 APM units for P1

(| p) (p e R) (R p a R l) (l “a R) (R e) (e u n) (n a l) (l “i) (“i N g) (g u) (u a) (a s t R) (R a n) (n j) (j “E R) (R a)

L1 to L2 EPM units for P1

(| p) (p e R) (R p a R) (R l a R) (R “E) (“E w) (w n “a l) (l i n) (n g) (g w) (w “a s) (s t R) (R a n) (n j) (j e R) (R “a)

L1 to L2 APM units for P2

(a | b i z) (z “o) (“o Gn a) (a i) (i m) (m p) (p a R) (R “a R) (R e l) (l e p) (p a R “O l) (l e d) (d i k) (k w) (w “e s) (s t a l) (l “i) (“i N) (N g) (g u) (u a) (a |)

L1 to L2 EPM units for P2

(“a | b i z) (z o n) (n j) (j “a) (“a i) (i m) (m p) (p a R) (R a R) (R “E l) (l “E p) (p a R) (R O) (O l “E) (“E d) (d “i k) (k w) (w e s) (s t) (t “a l) (l i) (i n) (n g) (g) (g w “a) (“a |)

Table 15-6: Comparison between unit selection using APM and, respectively, EPM for synthesizing the sentence: “Per parlare una lingua straniera, bisogna imparare le parole di questa lingua” (In order to speak a foreign language, one needs to learn the words). L1 French male voice; L2 Italian language; P1: phrase 1; P2: phrase 2. (---): unit; |: prosodic boundary; bold underlined: APM units with acoustic and prosodic faults; bold: corresponding EPM units. In additiom, for the other mappings developed for EPM, the perceived global quality in most of the concatenations seems to be higher than for APM. As far as unit length is concerned, in terms of number of consecutive phonemes a slight average increase is obtained by EPM over APM, as shown in Fig. 15-2, where the mean number of phonemes per unit, calculated for all 17 sentences and short paragraphs of the corpus, is reported for EPM and APM mappings respectively.

290

Chapter Fifteen

Fig. 15-2: Mean number of phonemes per unit for APM and, respectively, EPM, calculated over all the sentences generated and analyzed for this experiment (see the Appendix).

4. Subjective evaluation In order to assess the improvements obtained, a subjective evaluation test has been carried out by using the Pair Comparison methodology, a classic preference test, where listeners are asked to choose which of a pair of stimuli they prefer.9 A group of 10 Italian listeners (5 male and 5 female), familiar with speech synthesis, but not aware of the experiment in progress, have been subjected, through headphones, to 15 pairs of synthesized sentences/short paragraphs out of 17 of the entire corpus (13 French to Italian and 2 German to Italian mappings, see the Appendix) for evaluation by filling out the modified Pair Comparison questionnaire shown in Table 15-7. In each pair of sentences to be listened to and evaluated, one sentence had been generated by APM, the other by EPM. Moreover, in each pair the presentation order had been randomized. Listeners have been asked to accomplish the task from a workstation on which a dedicated on-line application had been installed. They were not constrained to time limits for giving their preference. The programme enabled them to listen to each sample as many times as they wanted. Then, listeners had to fill in the 9

Pier Luigi Salza et al., “MOS and Pair Comparison Combined Methods for Quality Evaluation of Text-to-Speech Systems”, Acta Acustica united with Acustica 82 (1996): 650-56.

Text-to-Speech Synthesis

291

electronic questionnaire before proceeding to the next sentence pair. On average, each test session lasted from 10 to 15 minutes. 1) Quale versione preferisci (Which version do you prefer)? x

E’ meglio la versione A (Version A is better than B)

x

E’ leggermente meglio la versione A (Version A is slightly better than B)

x

Sono praticamente uguali (They are nearly the same)

x

E’ leggermente meglio la versione B (Version B is slightly better than A)

x

E’ meglio la versione B (Version B is better than A)

Table 15-7: Modified Pair Comparison questionnaire. Results are presented in Fig. 15-3. Language dependent EPM generated samples were considered as sounding better or slightly better than APM generated samples in nearly 80% of cases, while another 7% of responses considered the two samples of equal quality. Although it has been carried out on a small number of synthetic samples by a limited number of subjects, this evaluation test represents an initial assessment of the possible improvements, in terms of acceptability and naturalness, that EPM can provide with respect to APM.

Fig. 15-3: Pair Comparison test results.

292

Chapter Fifteen

5. Conclusions The Automatic Phonetic Mapping (APM) enables each voice (L1) of Loquendo Text-to-Speech (LTTS®) to speak every language (L2) supported by the system, for mixed language text reading. The candidate phonemes are selected from L1’s phoneme set by means of a phonetic distance function based on weightings assigned to each articulatoryphonetic feature. APM has the advantage of being “universal”. However, in several cases the acoustic-prosodic quality of the concatenations generated by APM is not good. Fault analysis highlighted some critical aspects of APM, and suggested some ideas for managing the replacement of the APM output candidate phonemes when needed (repair rules). An Experimental Phonetic Mapping (EPM) prototype has been developed for working off-line, following some heuristic principles. In APM, the rigid definition of the weightings assigned by the phonetic distance function to each articulatory-phonetic feature isn’t suitable for every language. Thus, rules have been defined in EPM, for choosing the L1 phoneme, based on specific phonetic knowledge of both L1 and L2 languages. In APM, phonetic distance evaluation is strictly limited to the phoneme-by-phoneme criterion and does not account for the frequency of occurrences of the candidate phoneme within the L1 acoustic database. In EPM, the frequency of occurrences has been instead accounted for and the similarity has been evaluated within a larger phonetic context in order to favor the selection of longer units. By following these principles, examples of French to Italian and German to Italian mappings have been generated and analyzed, and comparisons of APM vs. EPM phonetic transcription and unit selection have been discussed. The perceived global quality of most concatenations generated by EPM seems to be higher than for APM. The trade-off between phonetic feature similarity and frequency of occurrence of phonetic sequences in the L1 speech database gives rise to an approximate but fluent, intelligible and plausible pronunciation of L2, consistent with what occurs in real speech: an L1 speaker tries to pronounce L2 “insertions” by maintaining both his own phonological system and his own prosody. The small subjective evaluation test carried out assesses the improvements obtained. Although the drawback of the present EPM prototype is that the repair rules have to be defined language by language, the experiment indicates interesting areas for investigation concerning naturalness and acceptability in making a TTS L1 voice speak an L2 language.

Text-to-Speech Synthesis

293

APPENDIX Texts of the sentences and short paragraphs used in the EPM mapping experiment (1 to 17) and in the subjective evaluation experiment (1 to 15). French male voice to Italian mapping 1. Per parlare una lingua straniera, bisogna imparare le parole di questa lingua (In order to speak a foreign language, it is important to learn the words). 2. Ma spesso succede che le persone non ti capiscono comunque (Yet, it often happens that people cannot understand you anyway). 3. Il fatto è, che non basta conoscere le parole, bisogna anche pronunciarle bene (As a matter of fact, it isn’t enough to know the words, we must also speak them well). 4. Altre lingue hanno suoni diversi, forse ne hanno di più. (Different languages have different sounds, maybe more). 5. Per esempio, il Giapponese ha in totale 110 suoni diversi, mentre l'Inglese ne ha oltre 8000 (For instance, Japanese has a total of 110 different sounds, whilst English has more than 8000). 6. Ecco perché i Giapponesi fanno fatica a pronunciare bene l'Inglese (That’s why the Japanese have difficulty in pronouncing English correctly). 7. Ma anche gli inglesi sono in difficoltà: non riescono a dare a ogni vocale lo stesso suono, come invece fanno i giapponesi (But even the British are in trouble: they can’t give each vowel the same sound, as the Japanese do). 8. Sono molto orgoglioso del risultato che abbiamo ottenuto quest'anno con la Ferrari, nonostante il fatto che entrambe le vetture non siano finite nei punti (I am very proud of what we have achieved this year with Ferrari, despite the fact that both cars finished with no points).

294

Chapter Fifteen

9. E’ vero, non hanno nemmeno terminato la gara, ma abbiamo comunque fatto un grande lavoro di squadra, specialmente i meccanici (It’s true, they didn’t even finish the race, but we still made a great team effort, especially the mechanics). 10. Vinceremo certamente l’anno prossimo, oppure nell'anno 2020 (We certainly will win next year or in 2020). 11. 14 marzo 2005. Roma. Giovanni Paolo II è tornato ieri a casa, intorno alle 18, questa volta non a bordo della papamobile, ma di una monovolume grigia. Il Papa, sorridente, ha salutato e benedetto i fedeli (March 14 2005. Rome. John Paul II returned home yesterday, at around 6 p.m., not aboard the Popemobile this time, but in a grey minivan. The Pope, smiling, greeted and blessed the faithful). 12. Prima di lasciare il policlinico "Gemelli", dove era ricoverato dallo scorso 24 febbraio, Wojtyla ha ringraziato personalmente i medici che lo hanno seguito in questi 18 giorni di degenza e non ha mancato di far sentire il suo affetto ai piccoli pazienti del reparto di oncologia pediatrica, ricoverati in un'ala adiacente al suo appartamento al decimo piano del policlinico (Before leaving the "Gemelli" hospital, where he had been recovering since last February 24, Wojtyla personally thanked the doctors who treated him during the 18 days of hospitalization. The Pope did not fail to show his affection for the young patients in the pediatric oncology department, in a wing adjacent to his room on the tenth floor of the hospital). 13. Il Pontefice ha dato appuntamento ai giovani per domenica prossima a San Pietro (The Pope has made an appointment for next Sunday with young people in St. Peter’s). German male voice to Italian mapping 14. Per parlare una lingua straniera, bisogna imparare le parole di questa lingua. Ma spesso succede che le persone non ti capiscono comunque. Il fatto è, che non basta conoscere le parole, bisogna anche pronunciarle bene (In order to speak a foreign language, it is important to learn the words. Yet, it often happens that people cannot understand you anyway. As a matter of fact, it isn’t enough to know the words, we must also speak them well).

Text-to-Speech Synthesis

295

15. Le lingue hanno suoni diversi, e hanno un numero diverso di suoni. Per esempio, il giapponese ha in totale 110 suoni diversi, mentre l'inglese, in base alla mia esperienza, ne ha oltre 8000. Ecco perché i Giapponesi fanno fatica a pronunciare bene l'inglese. Ma anche gli inglesi sono in difficoltà: non riescono mai a pronunciare ciascuna vocale con lo stesso suono, come invece fanno i giapponesi (Languages have different sounds, and a different number of sounds. For instance, Japanese has a total of 110 different sounds, whilst English, in my experience, has more than 8000. That’s why the Japanese have difficulty in pronouncing English well. But even the British are in trouble: they don’t give each vowel the same sound, as the Japanese do) Italian female voice to English 16. In order to speak a foreign language, it is important to learn the words. Yet it happens that people cannot understand you, even if the words are correct. As a matter of fact, it’s equally important to pronounce them correctly. Languages have different sounds, never forget that. Italian female voice to French 17. Voicì une exemple d'une phrase simple, prononcé avec une voix italienne. Vive la France, liberté, egalité, fraternité (Here is an example of a simple sentence, pronounced with an Italian voice. Vive la France, freedom, equality, fraternity).

REFERENCES

Achiba, Machiko. Learning to request in a second language. Child interlanguage pragmatics. Clevedon: Multilingual Matters, 2003. Albano Leoni, Federico. “Sulla voce.” In La voce come bene culturale, edited by Amedeo De Dominicis, 39-62. Roma: Carocci, 2002. ʊ. Dei suoni e dei sensi. Il volto fonico delle parole. Roma: Carocci, 2009. Albano Leoni, Federico, and Pietro Maturi. “Per una verifica pragmatica dei modelli fonologici.” In La linguistica pragmatica - Atti del XXIV Congresso della SLI, edited by Giovanni Gobber, 39-49. Roma: Bulzoni, 1992. Allen, James F., and Core Mark G. DAMSL: Dialog act markup in several layers (draft 2.1), Technical report, Multiparty Discourse Group, Discourse Research Initiative, September/October, 1997. Allwood, Jens. “Notes on Dialog and Cooperation.” In Proceedings of the IJCAI 97 Workshop on Collaboration, Cooperation and Conflict in Dialogue Systems, edited by Kristina Jokinen, David Sadek, and David Traum, 9–21. Nagoya, Japan, 1997. Anderson, Anne, Miles Bader, Ellen Bard, Elizabeth Boyle, Gwyneth Doherty, Simon Garrod, Stephen Isard, Jacqueline Kowtko, Jan McAllister, Jim Miller, Catherine Sotillo, Henry Thompson, and Regina Weinert. “The HCRC Map Task Corpus.” Language and Speech 34 (1991): 351-66. Anderson-Hsieh, Janet, Ruth Johnson, and Kenneth Koehler. “The Relationship Between Native Speaker Judgments of Nonnative Pronunciation and Deviance in Segmentals, Prosody, and Syllable Structure.” Language Learning 42/4 (1992): 529–55. Andorno, Cecilia. “Prima parla poi pensa: successo di una strategia di acquisizione basata sulla copia in Italiano L2.” Studi italiani di Linguistica teorica e applicata XXV/2 (1996): 291-311. ʊ. Focalizzatori fra connessione e messa a fuoco. Il punto di vista delle varietà di apprendimento. Milano: Franco Angeli, 2000. Anolli, Luigi M., and Rita Ciceri. La voce delle emozioni. Verso una semiosi della comunicazione vocale non-verbale delle emozioni. Milano: Franco Angeli, 1992.

298

References

Aoyama, Katsura, and Susan G. Guion. “Prosody in second language acquisition: Acoustic Analyses of duration and F0 range.” In Language experience in second language speech learning: In honor of James Emil Flege, edited by Oke-Schwen Bohn, and Murray J. Munro, 28297. Amsterdam: John Benjamins, 2007. Aureli, Massimo, Michela Cennamo, Claudia Crocco, and Francesca M. Dovetto. “Aspetti morfosintattici.” In Italiano parlato. Analisi di un dialogo, edited by Federico Albano Leoni, and Rosa Giordano, 119-35. Napoli: Liguori, 2005. Austin, John L. How to do things with words. Oxford, UK: Oxford University Press, 1962. Avesani, Cinzia, and Mario Vayra. “Costruzioni marcate e non marcate in italiano. Il ruolo dell’intonazione.” In Il parlante e la sua lingua. Atti delle X giornate del Gruppo di Fonetica Sperimentale, edited by Donatella Locchi, 1-14. Napoli: Istituto Universitario Orientale, 2000. Avesani, Cinzia, Mario Vayra, Daniela Banelli, and Maria Grazia Lovisolo. “Accentazione, deaccentazione e struttura informativa.” In Misura di parametri. Aspetti tecnologici ed implicazioni nei modelli linguistici. Atti del Primo Convegno Nazionale AISV, edited by Piero Cosi, 287-312. Brescia: EDK Editore, 2005. Bada, Erdo÷an. “Repetitions as vocalized fillers and self-repairs in English and French interlanguages.” Journal of Pragmatics 42/6 (2010): 168088. Badino, Leonardo, Claudia Barolo, and Silvia Quazza. “A general approach to TTS reading of mixed-language texts.” Paper presented at ICSLP 2004, Jeju, Korea, October 4-8, 2004. Bahler, Carly L., Coughlin Caitlin E., and Annie Tremblay. “Differential contribution of prosodic cues in native and non-native speech segmentation.” In Proceedings of the 17th International Congress of Phonetic Sciences, China 17-21 August 2011, edited by Wai-Sum Lee, and Eric Zee, 276-79. Hong Kong: City University of Hong Kong, 2011. Balboni, Paolo. Parole comuni, culture diverse. Guida alla comunicazione interculturale. Venezia: Marsilio, 1999. Balestri, Marcello, Alberto Pacchiotti, Silvia Quazza, Pier Luigi Salza, and Stefano Sandri. “Choose the best to modify the least: a new generation concatenative synthesis system.” Paper presented at EUROSPEECH ‘99, Budapest, Hungary, September 5-9, 1999. Barni, Monica, and Andrea Villarini. La questione della lingua per gli immigrati stranieri: insegnare, valutare e certificare l’italiano L2. Milano: Franco Angeli, 2001.

Prosodic and Rhythmic Aspects of L2 Acquisition: The Case of Italian

299

Barki Coricelli, Pazit, and Pierangela Diadori. Pro e contro. Conversare e argomentare in italiano. Livello intermedio. Guida per l’insegnante. Roma: Bonacci, 1997. Barry, William. “Rhythm as an L2 problem: How prosodic is it?” In NonNative Prosody. Phonetic Description and Teaching Practice, edited by Jürgen Trouvain, and Ulrike Gut, 97-120. Berlin / New York: Mouton de Gruyter, 2007. Barth-Weingarten, Dagmar, Nicole Dehé, and Anne Wichmann. Where Prosody meets Pragmatics. Studies in Pragmatics 8. Bingley: Emerald, 2009. Bazzanella, Carla. “Aspetti pragmatici della ripetizione dialogica”. In La linguistica pragmatica. Atti del XXIV Congresso della SLI, edited by Giovanni Gobber, 433-54. Roma: Bulzoni, 1992. ʊ. Le facce del parlare: un approccio pragmatico all’italiano parlato. Firenze: La Nuova Italia, 1994. ʊ. “Forme di ripetizione e processi di comprensione nella conversazione.” In La conversazione. Un’introduzione allo studio dell’interazione verbale, edited by Renata Galatolo, and Gabriele Pallotti, 205-25. Milano: Cortina, 1999. Bear, John, John Dowding, and Elizabeth. E. Shriberg. “Integrating multiple knowledge sources for detection and correction of repairs in human-computer dialog.” In Proceedings of the 30th Annual meeting of the association for computational linguistics, 56-63. Newark, Delaware, 1992. Beckman, Mary E., and Gayle M. Ayers. “Guidelines for ToBI Labelling. Version 3.0.” Unpublished manuscript, March 1997. Pdf file. Benucci, Antonella, ed. L’italiano libera-mente. L’insegnamento dell’italiano a stranieri in carcere. Perugia: Guerra, 2008. Berman, Ruth A., and Dan I. Slobin. Relating events in Narrative: a crosslinguistic developmental study. Hillsdale: Lawrence Erlbaum Associates Publishers, 1994. Bernini, Giuliano. “La banca dati del ‘Progetto di Pavia’ sull’italiano lingua seconda.” Studi Italiani di Linguistica Teorica e Applicata XXIII/2 (1994): 221–36. Berretta, Monica. “Tracce di coniugazione oggettiva in italiano.” In L’italiano tra le lingue romanze, edited by Fabio Foresti, Elena Rizzi, and Paola Benedini, 125-50. Roma: Bulzoni, 1989. ʊ. “Marcatezza in morfologia e apprendimento di lingue seconde.” Quaderni del Dipartimento di Linguistica e Letterature Comparate 8 (1992): 129-56.

300

References

ʊ. “Morphological markedness in L2 acquisition.” In Iconicity in Language, edited by Raffaele Simone, 197-233. Amsterdam: John Benjamins, 1995. Berruto, Gaetano. “Le dislocazioni a destra in italiano.” In Theme-Rheme in Italian, Thema-Rhema im Italienischen: symposium. Tema-Rema in Italiano, edited by Harro Stammerjoann, 55-69. Tübingen: Günter Narr Verlag, 1985. ʊ. “Per una caratterizzazione del parlato: l’italiano parlato ha un’altra grammatica?” in Gesprochenes Italienisch in Geschichte und Gegenwart, edited by Günter H. Holtus, and Edgar Radtke, 120-53. Tübingen: Günter Narr Verlag, 1985. ʊ. Sociolinguistica dell’italiano contemporaneo. Roma: La Nuova Italia Scientifica, 1987. ʊ. “Varietà diamesiche, diafasiche, diastratiche.” In Introduzione all’italiano contemporaneo. La variazione e gli usi, edited by Alberto Sobrero, 37-92. Roma: Laterza, 1993. Bertinetto, Pier Marco. Strutture prosodiche dell’italiano. Accento, quantità, sillaba, giuntura, fondamenti metrici. Firenze: Accademia della Crusca, 1981. Bertinetto, Pier Marco, and Chiara Bertini. “On modeling the rhythm of natural languages.” Paper presented at the biennal meeting for the Speech Prosody, Campinas, São Paulo, May 6-9, 2008. Bertinetto, Pier Marco, and Emanuela Magno Caldognetto. “Ritmo e intonazione.” In Introduzione all’italiano contemporaneo. Le strutture, edited by Alberto Sobrero, 141-92. Roma / Bari: Laterza & Figli, 1993. Bertinetto, Pier Marco, and Enrico Vivalda. “Recherches sur la perception des oppositions de quantité en italien.” Journal of Italian Linguistics 3 (1978): 97-116. Best, Catherine T., Gerald W. McRoberts, and Elizabeth Goodell. “Discrimination of non-native consonant contrasts varying in perceptual assimilation to the listener’s native phonological system.” Journal of the Acoustical Society of America 109 (2001): 775-94. Birch, Dawn, and Janelle McPhail. “The impact of accented speech in International television advertisements.” Global Business Languages 2/9 (2010): 91-105. Blanche-Benveniste, Claire. “La naissance des syntagmes dans les hésitations et répétitions du parler.” In Le sens et la mesure. Hommages à Benoît de Cornulier, edited by Jean Louis Araoui, 40-55. Paris: Honoré Champion, 2003.

Prosodic and Rhythmic Aspects of L2 Acquisition: The Case of Italian

301

Blum-Kulka, Shoshana, Juliane House, and Gabriele Kasper. Crosscultural pragmatics: Requests and apologies. Norwood, NJ: Ablex, 1989. Bocci, Giuliano. Aspetti teorici e analisi sperimentale della focalizzazione contrastiva nella periferia sinistra della frase: sintassi, fonologia e fonetica. Tesi di laurea non pubblicata, Siena: Facoltà di Lettere e Filosofia, 2003-2004. Boersma, Paul, and David Weenink. “Praat: doing phonetics by computer.” Computer program, version 5.3.10, March 12, 2012. http://www.praat.org/. Bonetti, Luka, Vlatka Utoviü, and Adinda Dulþiü. “Prosody in Speech Produced by Deaf Persons.” Croatian Review of Rehabilitation Research 44 (2008): 1-13. Bonomi, Ilaria. “La lingua dei quotidiani.” In La lingua italiana e i mass media, edited by Ilaria Bonomi, Andrea Masini, and Silvia Morgana, 127-64. Roma: Carocci, 2003. Boone, Daniel R. “Modification of the voices of deaf children.” The Volta Review 68 (1966): 686–92. ʊ. The voice and voice therapy. Englewood Cliffs, NJ: Prentice-Hall, 1971. Bou Franch, Patricia. “On Pragmatic Transfer.” Studies on English Language and Linguistics (1998): 5-18. Boula de Mareüil, Philippe, Giovanna Marotta, and Martine Adda-Decker. “Contribution of prosody to the perception of Spanish/Italian accents.” In Proceedings of the 2nd International Conference on Speech Prosody, Nara, Japan, 23-26 March 2004, edited by Bel Bernard, and Isabelle Marlien, 681-84. Nara 2004. Boula de Mareüil, Philippe, and Bianca Vieru-Dimulescu. “The contribution of prosody to the perception of foreign accent.” Phonetica 63 (2006): 247–67. Boulakia, Georges. “Linguistica e fonetica: senza voce o mezza voce?” In La voce come bene culturale, edited by Amedeo De Dominicis, 63124. Roma: Carocci, 2002. Bozzone Costa, Rosella, Luisa Fumagalli, and Ada Valentini, eds. Apprendere l’italiano da lingue lontane: prospettiva linguistica, pragmatica, educativa. Perugia: Guerra, 2012. Buroni, Edoardo. “La voce del telegiornale. Aspetti prosodici del parlato telegiornalistico italiano in chiave diacronica.” In L’italiano televisivo 1976-2006, Atti del Convegno Milano, 15-16 giugno 2009, edited by Elisabetta Mauroni, and Mauro Piotti, 387- 408. Firenze: Accademia della Crusca, 2010.

302

References

Bygate, Martin. “Quality of language and purpose of task: pattern of learners’ language on two oral communication tasks.” Language Teaching Research 3/3 (1999): 185-241. Calvert, D. R. “Deaf voice quality: A preliminary investigation.” The Volta Review 64 (1962): 401–3. Carletta, Jean, Stephen Isard, Gwyneth Doherty-Sneddon, Amy Isard, Jacqueline Kowtko, and Anne Anderson. HCRC Dialogue Structure Coding Manual. Technical Report, 82, Human Communication Research Centre, University of Edinburgh, 1996. Carletta, Jean, Amy Isard, Stephen Isard, Jacqueline Kowtko, Gwnyeth Doherty-Sneddon, and Anne Anderson. “HCRC Dialogue Structure Coding Manual.” Unpublished manuscript, June 3, 1996. Pdf file. Carlson, Lauri. Dialogue Games: An Approach to Discourse Analysis. Dordrecht, Holland: D. Reidel Co, 1983. Carroll, Mary. “Spatial information in descriptions: how advanced L2 speakers (L1 English - L2 German) manage coherence.” In Studi Italiani di Linguistica Teorica e Applicata, Strutture testuali e principi di organizzazione dell’informazione nell’apprendimento linguistico, edited by Marina Chini, and Anna Giacalone Ramat, 183-200. Roma: Pacini Editore, 1998. Caspers, Johanneke. “Who’s Next? The Melodic Marking of Question vs. Continuation in Dutch.” Language and Speech 41/3-4 (1998): 375-99. Celata, Chiara. Acquisizione e mutamento di categorie fonologiche: le affricate in italiano. Milano: Franco Angeli, 2004. ʊ. “The impact of allophonic variation on L2 speech perception.” In Recent Research in Second Language Phonetics/Phonology: Perception and Production, edited by Michael A. Watkins, Andreia S. Rauber, and Barbara O. Baptista, 64-80. Newcastle-upon-Tyne: Cambridge Scholars Publishing, 2009. Celata, Chiara, and Jessica Cancila. “Erosione sociolinguistica nell’italiano di emigrati. Il caso della lunghezza consonantica.” In Italiano: Acquisizione e perdita, edited by Stefania Scaglione, and Lidia Costamagna, 167-89. Milano: Franco Angeli, 2008. ʊ. “Phonological attrition and the perception of geminate consonants in the Lucchese community of San Francisco (CA).” International Journal of Bilingualism 14/2 (2010): 1-25. Cenoz, Jasone. “Pauses and hesitation phenomena in second language production.” ITL: Review of applied linguistics 127/128 (2000): 53-69. Cerrato, Loredana, and Mauro Falcone. “Acoustic and perceptual characteristic of Italian stop consonants.” Paper presented at the 5th

Prosodic and Rhythmic Aspects of L2 Acquisition: The Case of Italian

303

International Conference on Spoken Language Processing (ICSLP), Sydney, Australia, November 30 - December 4, 1998. Paper 0463. Chafe, Wallace. “Givenness, contrastiveness, definiteness, subjects, topics, and point of view.” In Subject and Topic, edited by Charles N. Li., 2556. London / New York: Academic Press, 1976. Chen, Aoju. “Language-specificity in the perception of continuation intonation.” In Tones and tunes II: phonetics and behavioural studies in word and sentence prosody, edited by Charles Gussenhoven, and Tomas Riad, 107-42. Berlin: Mouton de Gruyter, 2007. Chen, Sally, and Janice Fon. “Prosodic features of non-native English production.” In Proceedings of the 17th International Congress of Phonetic Sciences, China 17-21 August 2011, edited by Wai-Sum Lee, and Eric Zee, 456-59. Hong Kong: City University of Hong Kong. Chini, Marina, “Testualità e mezzi referenziali concernenti la persona in narrazioni di italofoni e di apprendenti di italiano L2.” In Studi Italiani di Linguistica Teorica e Applicata. Strutture testuali e principi di organizzazione dell’informazione nell’apprendimento linguistico, edited by Marina Chini, and Anna Giacalone Ramat, 153-81. Roma: Pacini Editore, 1998. ʊ. Che cos’è la linguistica acquisizionale. Roma: Carocci, 2005. Chini, Marina, and Anna Giacalone Ramat, eds. Topic, struttura dell’ informazione e acquisizione linguistica. Milano: Franco Angeli, 2010. Chun, Dorothy M. “Signal analysis software for teaching discourse intonation.” Language Learning and Technology 2/1 (1998): 61-77. ʊ. Discourse intonation in L2: From theory and research to practice. Amsterdam: John Benjamins, 2002. Chun, Dorothy M., Debra M. Hardison, and Martha C. Pennington. “Technologies for prosody in context: Past and future of L2 research and practice.” In Phonology and second language acquisition, edited by Jette G. Edwards Hansen, and Mary L. Zampini, 323-46. Amsterdam: John Benjamins, 2008. Ciliberti, Anna. “Repetition in native/non-native interaction.” In Repetition in dialogue, edited by Carla Bazzanella, 39-49. Tübingen: Niemeyer, 1996. Cohen, Andrew D. “Developing the ability to perform speech acts.” Studies in Second Language Acquisition 18 (1996a): 253-67. ʊ “Speech acts.” In Sociolinguistics and language teaching, edited by Sandra Lee McKay, and Nancy Hornberger, 383-420. New York: Cambridge University Press, 1996b. Cole, Jennifer, Mark Hasegawa-Johnson, Chilin Shin, Heejin Kim, EunKyung Lee., Hsin-yi Lu H. Yoonsook Mo, and Tae-Jin Yoon.

304

References

“Prosodic parallelism as a cue to repetition disfluency.” In Proceedings of DISS’05, Disfluency in spontaneous speech workshop, 1-4. Aix-enProvence, France 2005. Collier, René. “On the phonology of Dutch intonation.” In Worlds behind words: Essays in Honour of Professor F.G. Droste on the Occasion of his Sixtieth Birthday, edited by Frans J. Heyvaert, and Frieda Steurs, 245-58. Leuven: Leuven University Press, 1989. Collier, René, and Johan ‘t Hart. Cursus Nederlandse Intonatie. Leuven: Acco, 1981. Common European Framework of Reference for Languages, http://www.coe.int/t/dg4/linguistic/CADRE_EN.asp, accessed March 2012. Cook, Jenny, and John Gumperz. “Introduction.” In The social construction of literacy, edited by Jenny Cook, and John Gumperz. Cambridge: Cambridge University Press, 1986. Costamagna, Lidia. “The acquisition of Italian L2 affricates: the case of a Brazilian learner.” In New Sounds 2007: Proceedings of the Fifth International Symposium on the Acquisition of Second Language Speech, edited by Andreia S. Rauber, Michael A. Watkins, and Barbara O. Baptista, 138-48. Florianópolis: Federal University of Santa Catarina, 2007. http://www.nupffale.ufsc.br/newsounds/proceedings, ISBN: 978-8598703-04-6. ʊ. “L’apprendimento della fonologia dell’italiano da parte di studenti sinofoni: criticità e strategie.” In La didattica dell’italiano a studenti cinesi: il programma Marco Polo ed altre esperienze. Atti del XV seminario AICLU, edited by Elisabetta Bonvino, and Stefano Rastelli, 49-66. Pavia: Pavia University Press, 2011. Costamagna, Lidia, and Stefania Giannini, eds. La fonologia dell’interlingua: principi e metodi di analisi. Milano: Franco Angeli, 2003. Costamagna, Lidia, and Giovanna Marotta, eds. Acquisizione linguistica e categorie fonologiche. Pisa: Pacini, 2008. Costamagna, Lidia, and Cristina Montilli. “Le affricate italiane: percorsi di acquisizione.” In Acquisizione linguistica e categorie fonologiche, edited by Lidia Costamagna, and Giovanna Marotta, 67-92. Pisa: Pacini, 2008. Costamagna, Lidia, and Stefania Scaglione, eds. Italiano: Acquisizione e perdita. Milano: Franco Angeli, 2008. Cresti, Emanuela. Corpus di italiano parlato. Firenze: Accademia della Crusca, 2000.

Prosodic and Rhythmic Aspects of L2 Acquisition: The Case of Italian

305

Cresti, Emanuela, Philippe Martin, and Massimo Moneglia. “L’intonation des illocutions naturelles représentatives; analyse et validation perceptive.” In Macro-Syntaxe et Pragmatique. L’analyse linguistique de l’oral, edited by Mathieu Avanzi, 243-94. Roma: Bulzoni editore, 2003. Cresti, Emanuela, and Massimo Moneglia, eds. C-ORAL-ROM Integrated Reference Corpora for Spoken Romance Languages Studies, University of Florence, in Corpus Linguistics, 15. Amsterdam: John Benjamin Publishing Company, 2005. Crocco, Claudia. “Prosodic and informational aspects of polar questions in Neapolitan Italian.” In Proceedings of Speech Prosody, edited by Rudiger Hoffmann, and Hansjorg Mixdorff, 807-10. Dresden: TUC press, 2006. ʊ. “Topic accent and prosodic structure.” In Information structure and its interfaces, edited by Lunella Mereu, 15-49. Berlin / New York: Mouton de Gruyter, 2008. Cruttenden, Alan. “Falls and rises: meaning and universals.” Journal of Linguistics 17 (1981): 77-91. Curl, Traci S. “Repetition’ repairs: The relationship of phonetic structure and sequence organization.” In Sound Patterns in Interaction: Crosslinguistic studies from conversation, edited by Couper-Kuhlen, Elizabeth and Cecilia E. Ford, 273–98. Amsterdam: John Benjamins, 2004. ʊ. “Practices in other-initiated repair resolution: the phonetic differentiation of repetitions.” Discourse processes 39 (2005): 1-44. Curl, Traci S., John Local, and Gareth Walker. “Repetition and the prosody-pragmatics interface.” Journal of Pragmatics 38 (2006): 1721-51. Cutugno, Francesco, and Leandro D’Anna. “Segmenting the speech chains into tone units human behaviour versus automatic processes.” In Proceedings of the 15th International Congress of Phonetic Sciences, Barcelona, Spain 3-9 August 2003, edited by Solé Maria Josep, Recasens Daniel, and Joaquín Romero, 1233-36. Barcelona, Universitat Autonòma de Barcelona, 2003. Cutugno, Francesco, and Renata Savy. “Correlation between segmental reduction and prosodic features in spontaneous speech: the role of tempo.” In Proceedings of XIV International Conference of Phonetic Sciences (ICPhS), 471-74. San Francisco, 1999. Czigler, Peter E. “Timing in Swedish VC(C) Sequences.” Reports from the Department of Phonetics Umeå University, PHONUM 5, Umeå, Sweden: Department of Phonetics, Umeå University, 1998.

306

References

D’Achille, Paolo. “Attraverso i ‘ponti’ dell’Abruzzo e del Molise.” Italiano e Oltre XI/5 (1996): 285-91. De Beaugrande, Robert A., and Wofgang U. Dressler. Introduction to Text Linguistics. London: Longman, 1981. De Bot, Kees. “Visual feedback of intonation I: Effectiveness and induced practice behavior.” Language and Speech 26/4 (1983): 331-50. ʊ. “The transfer of intonation and the missing database.” In Crosslinguistic Influence in Second Language Acquisition, edited by Eric Kellerman, and Michael Sharwood Smith, 110-19. New York: Pergamon Press, 1986. De Dominicis, Amedeo. Intonazione e contesto. Alessandria: Edizioni dell’Orso, 1992. ʊ. “La voce come bene culturale, la linguistica come strumento di catalogazione e restauro.” In La voce come bene culturale, edited by Amedeo De Dominicis, 7-26. Roma: Carocci, 2002. ʊ. Intonazione. Una teoria della costituenza delle unità intonative. Roma: Carocci, 2010. De Mauro, Tullio. Storia linguistica dell’Italia unita. Roma / Bari: Laterza, 1963. De Mauro, Tullio, and Massimo Vedovelli. Italiano 2000. I pubblici e le motivazioni dell’italiano diffuso fra stranieri. Roma: Bulzoni, 2003. De Pietro, Jean François, Marinette Matthey, and Bernard Py. “Acquisition et contrat didactique: les séquences potentiellement acquisitionnelles dans la conversation exolingue.” In Actes du 3e colloque régional de linguistique, edited by Dominique Weil, and Huguette Fugier, 99-124. Strasbourg: Presses de l’Université de Sciences Humaines, 1989. de Pijper, Jan Roelof, and Angelien A. Sanderman. “On the perceptual strength of prosodic boundaries and its relation to suprasegmental cues.” Journal of Acoustical Society of America 96 (1994): 2037-47. Del Giudice, Alex, Shosted Ryan, Kathryn Davidson, Mohammad Salihie, and Amalia Arvaniti. “Comparing methods for locations pitch ‘elbows’.” In Proceedings of 16th ICPhS, edited by Jürgen Trouvain, and William Barry, 1117-20. Saarbrücken: Universität des Saarlandes, 2007. Delmonte, Rodolfo. “Prosodic tools for language learning.” International Journal of Speech Technology 4 (2009): 161-184. Derwing, Tracey M., and Murray J. Munro. “Accent, comprehensibility and intelligibility: Evidence from four L1s.” Studies in Second Language Acquisition 19 (1997): 1-16.

Prosodic and Rhythmic Aspects of L2 Acquisition: The Case of Italian

307

Derwing, Tracey M., Murray J. Munro, and Grace Elaine Wiebe. “Pronunciation instruction for ‘fossilized’ learners: Can it help?” Applied Language Learning 8 (1997): 217-35. Derwing, Tracey M., Marian J. Rossiter, Murray J. Munro, and Ron I. Thomson. “Second Language Fluency: Judgments on Different Tasks.” Language Learning 54 (2004): 655-79. DeShields, Oscar W., Jr., and Gilberto de los Santos. “Salesperson’s accent as a globalization issue.” Thunderbird International Business Review 42 (2000): 29-46. DeShields, Oscar W. Jr., Ali Kara, and Erdener Kaynak. “Source effects in purchase decisions: The impact of physical attractiveness and accent of salesperson.” International Journal of Research in Marketing 13 (1996): 89-101. Devís Herraiz, Empar. “La prosodia nell’interferenza tra L1 e L2: il caso delle interrogative polari tra veneti e catalani”. Estudios de fonética experimental 16 (2007): 119-46. Di Russo, Domenico. Intonazione e modalità nella significazione. PhD diss., Università di Roma “La Sapienza”, 2011. Diadori, Pierangela. L’italiano televisivo. Siena: Bonacci, 1994. Dörnyei, Zoltan, and Mary Lee Scott. “Communication strategies in a second language: definitions and taxonomies.” Language learning 47/1 (1997): 173-210. Duez, Danielle. “Silent and non-silent pauses in three speech styles.” Language and Speech 25 (1982): 11-28. ʊ. Le pause dans la parole de l’homme politique. Paris: Editions du Centre National de la Recherche Scientifique, 1991. Dunn, Malcom H. “The Phonetics and Phonology of Geminate Consonants: A Production Study.” Unpublished PhD diss., Yale University, 1993. Ehala, Martin. “Estonian quantity: Implications for Moraic Theory.” In Generative Approaches to Finnic and Saami Linguistics, edited by Diane Nelson, and Satu Manninen, 51-80. Stanford: CSLI, 2003. Ellis, Rod. Implicit and explicit learning of language. London: Academic Press, 1994. ʊ. The Study of Second Language Acquisition. Oxford: Oxford University Press, 1994. ʊ. Task-based Language Learning and Teaching. Oxford: Oxford University Press, 2003. Engstrand, Olle, and Diana Krull. “Durational correlates to quantity in Swedish, Finnish and Estonian: Cross-language evidence for a theory of adaptive dispersion.” In Festschrift on the Occasion of Björn

308

References

Lindblom’s 60th Birthday, edited by Olle Engstrand, and Klaus Kohler, Special issue of Phonetica, 51 (1994): 80-91. Esposito, Anna, and Maria Gabriella Di Benedetto. “Acoustic and perceptual study of gemination in Italian stops.” Journal of the Acoustical Society of America 30 (1999): 175-85. Ferreri, Silvana. “Italiano regionale di Sicilia: dove si parla e dove se ne parla.” Italiano e Oltre X/4 (1995): 44-45. Fetzer, Anita, and Kirstin Fischer eds. Lexical markers of common grounds. Amsterdam: Elsevier, 2006. Finkelstein, Yehuda, Jacob Bar-Ziv, Ariela Nachmani, Gilead Berg er, and Dov Ophir. “Peritonsillar abscess as a cause of transient velopharyngeal insufficiency.” Cleft Palate-Craniofacial Journal 30 (1993): 421–28. Fitzpatrick, Jennifer. “On intonational typology.” In Methodological issues in language typology, edited by Peter Siemund, Sprachtypologie und Universalien Forschung 53 (2000): 88-96. Flege, Jim E. “Assessing constraints on second-language segmental production and perception.” In Phonetics and Phonology in Language Comprehension and Production: Differences and Similarities, edited by Niels O. Schiller, and Antje S. Meyer, 319–55. Berlin: Mouton de Gruyter, 2003. Fonagy, Ivan. La vive voix. Essai de psycho-phonetique. Paris: Payot, 1983. Ford, Cecilia E., and Sandra A. Thompson. “Interactional units in conversation: syntactic, intonational, and pragmatic resources for the management of turns.” In Interaction and Grammar, edited by Elinor Ochs, Emanuel A. Schegloff, and Sandra A. Thompson, 134–84. Cambridge: Cambridge University Press, 1996. Francis, Caryn. “Talk to me! The development of request strategies in non-native speakers of English.” Working Papers in Educational Linguistics (1997): 23–40. Francuzik, Katarzyna, KarpiĔski Maciej, and Janusz Klesta. “A preliminary study of the Polish intonational phrase, nuclear melody and pauses in Polish semi-spontaneous narration.” In Proceedings of the 1st International Conference on Speech Prosody, Aix-en-Provence (France), 11-13 Avril 2002, edited by Bel Bernard, and Marlien Isabelle, 303-306. Aix-en Provence 2002, Laboratoire Parole et Langage. Frascarelli, Mara. “Dislocation, Clitic Resumption and Minimality: A Comparative Analysis of Left and Right Topic Constructions in Italian.” In Romance Languages and Linguistic Theory, edited by

Prosodic and Rhythmic Aspects of L2 Acquisition: The Case of Italian

309

Reineke Bok-Bennema, Bart Hollebrandse, Brigitte Kampers-Mahne, and Petra Sleeman, 99-118. Amsterdam / Philadelphia: John Benjamins, 2004. Galatà, Vincenzo, and Luciano Romito. “Un corpus sperimentale per lo studio cross-linguistico europeo delle emozioni vocali.” In Proceedings of the 5th AISV National Conference - La dimensione Temporale del parlato, Università di Zurigo, Svizzera, 4-6 February 2009, edited by Stephan Schmid, Michael Schwarzenbach, and Dieter Studer, 603-42. Torriana (RN): EDK, 2010. Gass, Susan M., and Larry Selinker. Second Language Acquisition: An Introductory Course. 2nd edition, Mahwah, NJ: Lawrence Erlbaum Associates, 2001. Genç, Bilal, Mustafa Mavaúo÷lu, and Erdo÷an Bada. “Types and functions of repetitions in the narrations of Turkish speakers of French.” Research on youth and language 4 (2010): 216-24. Germain, Aline, and Philippe Martin. “Présentation d’un logiciel de visualisation pour l’apprentissage de l’oral en langue seconde.” Apprentissage des Langues et Systèmes d’Information et de Communication 3/1 (2000): 61-76. Giacalone Ramat, Anna, ed. Verso l’italiano. Percorsi e strategie di acquisizione. Roma: Carocci, 2003. Giannini, Antonella. “Range di variabilità della velocità di articolazione in italiano.” In Atti del XXVIII Convegno Nazionale dell’AIA, 253-56. Trani, 2000. ʊ. “Analisi acustica del parlato televisivo.” In Misura dei Parametri, aspetti tecnologici ed implicazioni nei modelli linguistici. I convegno Nazionale dell’Associazione Italiana di Scienze della Voce (AISV), edited by Piero Cosi, 49-61. Padova: EDK, 2004. ʊ. “I silenzi del telegiornale.” In La Comunicazione Parlata, Atti del Congresso internazionale, Napoli 23-25 febbraio 2006, edited by Massimo Pettorino, Antonella Giannini, Marianna Vallone and Renata Savy, vol I, eBook. Napoli: Liguori editore, 2006. Giannini, Antonella, and Massimo Pettorino. “Il parlato dei mass media: analisi multilingue del parlato dei telegiornali.” In La Comunicazione Parlata 3, edited by Massimo Pettorino, Antonella Giannini, and Francesca Dovetto, vol. II. Napoli: Università degli Studi di Napoli “L’Orientale”, 2010. Giannini, Stefania, and Lidia Costamagna. “Acquisizione di categorie fonologiche e diffusione lessicale del mutamento linguistico: affinità strutturali.” Archivio Glottologico Italiano 83 (1998): 150-87.

310

References

Gili Fivela, Barbara, Cinzia Avesani, Giuliano Bocci, Mariapaola D’Imperio, Rosa Giordano, Giovanna Marotta, Michelina Savino, and Patrizia Sorianello. “Varieties of Italian and their ToBI transcription”. Paper presented at the Workshop on Romance ToBI (PaPI 2011), Terragona, June 23, 2011. Giordano, Rosa. “The intonation of polar questions in two central varieties of Italian.” In Proceedings of Speech Prosody, edited by Rudiger Hoffmann and Hansjorg Mixdorff, Dresden: TUC press, 2006: PS8-10155.pdf. Gìvon, Talmy. Discourse and syntax. New York: Academic Press, 1979. Goad, Heather, and Lydia White. “Ultimate attainment in interlanguage grammars: a prosodic approach.” Second Language Research 22/3 (2006): 243-68. Goldsmith, John A. “Autosegmental Phonology.” PhD diss., Massachusetts Institute of Technology, 1976. New York: Garland, 1979. Goodwin, Charles. “Restarts, pauses, and the achievement of mutual gaze at turn-beginning.” Sociological inquiry 50/3-4 (1980): 272–302. Gooskens, Charlotte, and Vincent J. van Heuven. “Declination in Dutch and Danish: Global versus local pitch movements in the perceptual characterisation of sentence types.” In Proceedings of ICPhS ’95, edited by Kjell Elenius, and Peter Branderud, 2: 374-77. Stockholm: Stockholm University, 1995. Gossen, Carl Th. Studien zur syntaktischen und stilistischen Hervorhebung im modernen Italienisch. Berlin: Akademie-Verlag, 1954. Grabe, Esther, and Low Ee Ling. “Durational variability in speech and the rhythm class hypothesis.” In Papers in Laboratory Phonology 7, edited by N. Warner, and Carl Gussenhoven, 515-46. Berlin: Mouton de Gruyter, 2002. Grabe, Esther, and Maciej KarpiĔski. “Universal and language-specific aspects of intonation in English and Polish.” In Proceedings of the 15th International Congress of Phonetic Sciences, Spain 3-9 August 2003, edited by Solé Maria Josep, Recasens Daniel, and Joaquín Romero, 14. Barcelona: Universitat Autonòma de Barcelona, 2003. Grassi, Corrado, Alberto Sobrero, and Tullio Telmon. Introduzione alla dialettologia italiana. Roma / Bari: Laterza, 2003. Grice, Martine, Ralf Benzmüller, Michelina Savino, and Bistra Andreeva. “The intonation of queries and checks across languages: data from Map Task dialogs.” In Proceedings of the XIII International Congress of Phonetic Sciences 3, edited by Kjell Elenius, and Peter Branderud, 648-51. Stockholm: Stockholm University, 1995.

Prosodic and Rhythmic Aspects of L2 Acquisition: The Case of Italian

311

Grice, Martine, Mariapaola D’Imperio, Michelina Savino, and Cinzia Avesani. “Intonation and communicative function in a regional variety of Italian.” Phonus 1 (1995): 19-22. ʊ. “Question type and information structure in Italian.” In Proceedings of Prosodic Interface, Nantes, France, edited by Amina Mettouchi, and Gaëlle Ferré, 117-22. Nantes: Université de Nantes, UFR Lettres et Langage, AAI, 2003a. ʊ. “Map Tasks in Italian: asking questions about given, accessible and new information.” Catalan Journal of Linguistics 2 (2003b): 153-80. ʊ. “Information structure and questions: evidence from task-oriented dialogues in a variety of Italian.” In Regional Variation in Intonation, edited by Peter Gilles and Jörg Peters, 161-87 V/240 S. Linguistische Arbeiten. Tübingen: Niemeyer, 2004. ʊ. “Strategies for intonation labelling across varieties of Italian.” In Prosodic Typology: the Phonology of Intonation and Phrasing, edited by Sun-Ah Jun, 362-389. New York: Oxford University Press, 2005. Grosser, Wolfgang. “Aspects of intonation L2 acquisition.” In Current Issues in European Second Language Acquisition Research, edited by Bernhard Kettemann, and Wilfried Wieden, 81-94. Tübingen: Gunter Narr Verlag, 1993. Gumperz, John. Discourse strategies. Cambridge: Cambridge University Press, 1982. Gussenhoven, Carlos. “Adequacy in intonation analysis: The case of Dutch.” In Autosegmental Studies on Pitch Accent, edited by Harry van der Hulst, and Norval Smith, 95-122. Dordrecht: Foris, 1988. ʊ. “Tone Segments in the Intonation of Dutch.” In The Berkeley Conference on Dutch Linguistics 1989, edited by Thomas F. Shannon, and Johan P. Snapper, 139-55. Lanham: University Press of America, 1991. ʊ. “Transcription of Dutch intonation.” In Prosodic Typology: The phonology of intonation and phrasing, edited by Sun-Ah Jun, 118-45. Oxford: Oxford University Press, 2005. Gussenhoven, Carlos, and Toni Rietveld. “A target-interpolation model for the intonation of Dutch.” In Proceedings of ICSLP ’92, edited by John J. Ohala, Terrance M. Nearey, Bruce L. Derwing, Megan M. Hodge, and Grace E. Wiebe, 1235-38. Banff: University of Alberta, 1992. Gut, Ulrike. “Prosody in second language speech production: the role of the native language.” Fremdsprachen Lehren und Lernen 32 (2003): 133-52.

312

References

Haan, Judith. Speaking of questions: an exploration of Dutch question intonation. Utrecht: Landelijke Onderzoekschool Taalwetenschap, 2002. Habtemariam, Marcos. “Italian.” In Language in Ethiopia, edited by Marvin L. Bender, Donald Bowen, Robert L. Cooper, and Charles A. Ferguson, 170-82. Oxford: Oxford University Press, 1976. Hall, Judit A. “Voice and tone persuasion.” Journal of Personality and Social Psychology 38/6 (1980): 924-34. Ham, William H. Phonetics and phonological aspects of geminate timing. London: Routledge, 2001. Hansen, Benjamin B. “Production of Persian Geminate Stops: Effects, of varying speaking rate.” In Proceedings of the 2003 Texas Linguistics Society Conference, edited by Augustine Agwele, Willis Warren, and Sang-Hoon Park, 86-95. Somerville, MA: Cascadilla Proceedings Project, 2004. Harlow, Linda L. “Do they mean what they say? Sociopragmatic competence and second language learners.” The Modern Language Journal 74 (1990): 328–49. ’t Hart, Johan, René Collier, and Antonie Cohen. A perceptual study of intonation. An experimental-phonetic approach to speech melody. Cambridge: Cambridge University Press, 1990. Hawkins, Sarah. “Auditory Capacities and Phonological Development: Animal, Baby and Foreign Listeners.” In The Acoustics of Speech Communication, edited by James M. Pickett, 183-198. Needham Heights, MA: Allyn and Bacon, 1999. Hayes, Bruce. “Compensatory lengthening in moraic phonology.” Linguistic Inquiry 20 (1989): 253-306. Hedberg, Nancy, Juan M. Sosa, and Lorna Fadden. “Meanings and Configurations of Questions in English.” In Proceedings of Speech Prosody, edited by Bernard Bel, and Isabelle Marlien. Nara, Japan, 2004. Hindle, Donald. “Deterministic parsing of syntactic non-fluencies.” In Proceedings of the 21st annual meeting of the association for computational linguistics, 123-28, 1983. Hirschberg, Julia. “Pragmatics and Intonation.” In The Handbook of Pragmatics, edited by Laurence R. Horn, and Gregory Ward, 515-37. Oxford: Blackwell Publishing, 2006. Hirst, Daniel, Albert Di Cristo and Robert Espesser. “Levels of representation and levels of analysis for intonation.” In Prosody: Theory and Experiment, edited by Merle Horne, 51-87. Dordrecht: Kluwer Academic Publishers, 2000.

Prosodic and Rhythmic Aspects of L2 Acquisition: The Case of Italian

313

Hovland, Carl I., Janis L. Irving, and Harold H. Kelley. Communication and Persuasion. New Haven: Yale UP, 1953. Hsieh, Fuhui “Repetition in social interaction: a case study on Mandarin Conversations.” International Journal on Asian language Processing 19 (2009): 153-68. Interlandi, Grazia M. “Il continuum della variazione pragmaticoespressiva nell’intonazione dell’italiano parlato a Torino.” In Il parlato italiano. Atti del Convegno nazionale, edited by Federico Albano Leoni, Francesco Cutugno, Massimo Pettorino, and Renata Savy, 1-20. Napoli: M. D’Auria Editore, CD-Rom, 2004. Iori, Beatrice and Sergio Neri, eds. L’italiano e le altre lingue, Apprendimento della seconda lingua e bilinguismo dei bambini e dei ragazzi immigrati. Milano: Franco Angeli, 2005. Jalilifar, Alireza. “Request Strategies: Cross-Sectional Study of Iranian EFL Learners and Australian Native Speakers.” English Language Teaching 2 (2009): 46-61. James, Eric. “The Acquisition of Prosodic Features of Speech Using a Speech Visualizer. IRAL 14 (1976): 227-44. Jassem, Wiktor. “Computer-based classification of basic Polish intonations.” In Proceedings of the 11th International Congress of Phonetic Sciences, Tallin (Estonia) August 1-7 1987, Estonia, USSR, 253-56. Jefferson, Gail. “On the poetics of ordinary talk.” Text and Performance Quarterly 16 (1996): 1–61. Jilka, Matthias. “The contribution of intonation to the perception of foreign accent. Identifying intonational deviations by means of F0 generation and resynthesis.” Ph.D. diss., University of Stuttgart, 2000. Jones, Caroline, Lynn Berry, and Catherine Stevens. “Synthesized speech intelligibility and persuasion: Speech rate and non-native listeners.” Computer Speech and Language 21 (2007): 641-51. Jönsson, Arne. “Dialogue Actions for Natural Language Interfaces.” Proceedings of IJCAI-95, edited by Chris Mellish, 1406-11. Montrèal, Canada: Morgan Kaufmann Publishers, 1995. Jun, Sun-Ah, and Mira Oh. “Acquisition of Second Language Intonation.” In Proceedings of ICSLP 2000, edited by Baozong Yuan, Taiyi Huang, and Xiaofang Tang, 4: 76-79. Beijing: China Military Friendship Publishers, 2000. Kaglik, Anna, and Philippe Boula de Mareüil. “Polish-accented French prosody in perception and production: transfer or universal acquisition process?” In Proceedings of the 5th International Conference on

314

References

Speech Prosody Chicago, Illinois 11-14 May, 2010, Illinois 2010, University of Illinois Press, available to http://speechprosody2010.illinois.edu/program.php/papers/100987.pdf. Kanouse, David E., and Robert P. Abelson. “Language variables affecting the persuasiveness of simple communication.” Journal of Personality and Social Psychology 7/2-1 (1967): 158-63. Karcevskij, Sergej. “Sur la Phonologie de la Phrase.” Travaux du Cercle Linguistique de Prague IV (1931): 188-227. KarpiĔski, Maciej. “The corpus of Polish intonational database. Technical specification.” In Investigationes Linguisticae VIII (2002): 7-8. KarpiĔski, Maciej, and Ewa J. Nowikow. “Prosodic and gestural features of phrase-internal disfluencies in Polish spontaneous utterances.” In Proceedings of the 5th International Conference on Speech Prosody Chicago Illinois 11-14 May, 2010, Illinois 2010, University of Illinois Press, available to http:// speechprosody2010.illinois.edu/program.php/papers/100152.pdf. Kasper, Gabriele. “Pragmatic Transfer.” Second Language Research 8 (1992): 203-31. ʊ. Can pragmatic competence be taught?, 1997, accessed March 2012, http://www.nflrc.hawaii.edu/NetWorks/NW06/. Kiparsky, Paul. “Livonian stød.” In Segments and Tone edited by Paul Boersma, Marc van Oostendorp, Ben Hermans, and,Wolfgang Kehrei, 1-16.Tübingen: Niemeyer, 2006. Klein, Wolfgang, and Clive Perdue. “The basic variety (or: couldn’t natural languages be simpler?).” Second Language Research 13 (1992): 301-47. Klein, Wolfgang, and Christian von Stutterheim. “Quaestio und referentielle Bewegung in Erzählungen.” Linguistische Berichte 109 (1987): 163-83. Kohler, Klaus J. “The production of plosives.” Arbeitsberichte des Instituts für Phonetik der Universität Kiel (AIPUK) 8 (1977): 30-110. Kormos, Judih. “Monitoring and Self-repair in L2.” Language learning 49/2 (1999): 303-42. Krull, Diana. “Foot isochrony in Estonian.” In Proceedings of the XIV International Congress of Phonetic Sciences (ICPhS ’99) edited by John J. Ohala, Yoko Hasegawa, Manjari Ohala, Daniel Granville, and Ashlee C. Bailey, 1063-66. San Francisco: University of California, 1999.

Prosodic and Rhythmic Aspects of L2 Acquisition: The Case of Italian

315

Kuhl, Patricia K. “Human adults and human infants show a ‘perceptual magnet effect’ for the prototypes of speech categories, monkeys do not.” Perception & Psychophysics 50 (1991): 93–107. Ladd, Robert D. Intonational phonology. Cambridge: Cambridge University Press, 1996 (1st ed.), 2008. Lambrecht, Knud L. Topic, antitopic and verb agreement in non-standard French. Amsterdam: John Benjamins, 1981. ʊ. Information structure and sentence form: Topic, focus, and the mental representation of discourse referents. Cambridge Studies in Linguistics 71, Cambridge: Cambridge University Press, 1994. Lane, Harlan, and Roger Buiten. “A Self-Instructional Device For Conditioning Accurate Prosody: Trends.” In Language Teaching, edited by Albert Valdman, 159-74. New York: Academic Press,1969. Lehiste, Ilse. Consonant quantity and phonological units in Estonian. Bloomington: Indiana University Publications, 1966. ʊ. “The search for phonetic correlates in Estonian prosody.” In Estonian Prosody: Proceedings of the International Symposium on Estonian Prosody, edited by Ilse Lehiste, and John Ross, 11-33. Tallinn: Institute of Estonian Language, 1997. Lehiste, Ilse, Katherine Morton, and Mark A. A. Tatham. “An instrumental study of consonant germination.” Journal of Phonetics 3 (1973): 131-48. Lenden, Jessica M., and Peter Flipsen Jr. “Prosody and voice characteristics of children with cochlear implants.” Journal of Communication Disorders 40 (2007): 66–81. Levelt, Willem. “Monitoring and self-repair in speech.” Cognition 14 (1983): 41-104. Levelt, Willem, and Anne Cutler. “Prosodic marking in speech repair.” Journal of semantics 2 (1983): 205-17. Levi, Susannah V., and David B. Pisoni. “Indexical and linguistic channels in speech perception: Some effects of voiceovers on advertising outcomes.” In Psycholinguistic Phenomena in Marketing Communications, edited by Tina M. Lowrey, 203-19. Mahwah, NJ: Lawrence Erlbaum Associates, 2007. Li, Deng, and Jianwu Dang. “Speech analysis: The production-perception perspective.” In Advances in Chinese Spoken Language Processing, edited by Chin-Hue Lee, Haizhou Li, Lin Shan Lee, Ren Hua Wang, and Qiang Huo, 2-32. Singapore: World Scientific Publishing, 2007. Lindblom, Björn. “Explaining phonetic variation: a sketch of the H&H theory.” In Speech production modelling, edited by William Harcastle, and Alain Marchal, 403-39. Dordrecht: Kluwer publ., 1990.

316

References

Lo Duca, Maria G. “Sulla rilevanza per la glottodidattica dei dati dell’acquisizione di lingue seconde: “narrare” in italiano L2.” In Verso l’Italiano, edited by Anna Giacalone Ramat, 254-70. Roma: Carocci, 2003. Loporcaro, Michele. “On the analysis of geminates in Standard Italian and Italian dialects.” In Natural Phonology: The State of the Art, edited by Bernhard Hurch, and Richard A. Rhodes, 153-87. Berlin: Mouton de Gruyter, 1992. Lyster, Roy. “Recasts, repetition, and ambiguity in L2 classroom discourse.” Studies in second language acquisition 20 (1998): 51-81. Magno Caldognetto, Emanuela, Federica Cavicchio, and Piero Cosi, eds. Comunicazione parlata e manifestazione delle emozioni. Atti del 1° Convegno Nazionale GSCP - Gruppo di Studio della Comunicazione Parlata. Padova 30 Novembre - 1 Dicembre 2004. Napoli: Liguori, 2008. Magno Caldognetto, Emanuela, Enrica De Zordi, and Loredana Corrà. “Il ruolo delle pause nella produzione della parola.” Il Valsalva Bollettino della Società Italiana di Audiologia e Foniatria 5 (1982): 12-21. Magno Caldognetto, Emanuela, Franco Ferrero, Clara Lavagnoli, and Kyriaki Vagges. “F0 contours statements, yes-no questions and whquestions of two regional varieties of Italian.” Journal of Italian Linguistics 3 (1978): 57-68. Magno Caldognetto, Emanuela, and Kiki Vagges. “Indici di fluenza, tipologia e distribuzione delle sillabe nel parlato spontaneo.” In Atti del XIX Convegno nazionale AIA, 423-29. Napoli, 1991. Maldonado, Rachel, Patriya Tanshuaj, and Darrel D. Muehling. “The impact of gender on Ad processing: A social identity perspective.” Academy of Marketing Science Review 3 (2003): 1-15. Marotta, Giovanna. Modelli e misure ritmiche. La durata vocalica in italiano. Bologna: Zanichelli, 1985. Marotta, Giovanna, and Patrizia Sorianello. “Question intonation in Sienese Italian.” In Proceedings of the XIV International Conference of Phonetic Sciences, edited by John J. Ohala, Yoko Hasegawa, Manjari Ohala, Daniel Granville and Ashlee C. Bailey, 1161-64. Berkeley: University of California, 1999. Martin, Philippe. “L’intonation de la phrase dans les langues romanes: l’exception du français.” Langue française mars 2004: 36-55. ʊ. Intonation du français. Paris: Armand Colin, 2009. ʊ. “Prominence detection without syllabic segmentation, Prosodic Prominence: Perceptual and Automatic Identification.” Paper

Prosodic and Rhythmic Aspects of L2 Acquisition: The Case of Italian

317

presented at Speech Prosody, 5th International Conference - Satellite Workshop, Chicago, Illinois, May 11-14, 2010. Maturi, Pietro. “L’intonazione delle frasi dichiarative ed interrogative nella varietà napoletana dell’italiano.” Rivista Italiana di Acustica 12 (1988): 13-30. McCrary, Kristie M. “Reassessing the Role of the Syllable in Italian Phonology: An Experimental Study of Consonant Cluster Syllabification, Definite Article Allomorphy and Segment duration.” Ph.D. diss., UCLA, 2003. McGarr, Nancy S., and Mary Joe Osberger. “Pitch deviancy and intelligibility of deaf speech.” Journal of Communication Disorders 11 (1978): 237–47. Meisel, Jürgen, Harold Clashen, and Manfred Pienemann. “On Determining Developmental Stages in Natural Second Language Acquisition.” Studies in Second Language Acquisition, 3/2 (1981): 109-35. Menn, Lise, and Suzanne Boyce. “Fundamental frequency and discourse structure.” Language and Speech 25 (1982): 341-83. Mennen, Ineke. “Second Language Acquisition of Intonation: The Case of Peak Alignment.” In CLS 34-2: The Panels, edited by M. Catherine Gruber, Derrick Higgins, Kenneth S. Olson, and Tamra Wysocki, 32741. Chicago: Chicago Linguistic Society, 1998. ʊ. The realization of nucleus placement in second language intonation.” In Proceedings of ICPhS’99, edited by John J. Ohala, 1: 555-58. San Francisco: University of California, 1999. ʊ. “Bi-directional interference in the intonation of Dutch speakers of Greek.” Journal of Phonetics 32/4 (2004): 543-63. ʊ. “Phonological and phonetic influences in non-native intonation.” In Non-Native Prosody. Phonetic Description and Teaching Practice, edited by Jürgen Trouvain, and Ulrike Gut, 53-76. Berlin / New York: Mouton de Gruyter, 2007. Mettouchi, Amina, Anne Lacheret-Dujour, Vered Silber-Varod, and Izre’el. Shlomo. “Only prosody? Perception of speech segmentation in Kabyle and Hebrew.” Nouveaux cahiers de linguistique française 28 (2007): 207-18. Minuz, Fernanda. Italiano L2 e alfabetizzazione in età adulta. Roma: Carocci, 2005. Mo, Yoonsook. “Duration and intensity as perceptual cues for naïve listeners’ prominence and boundary perception.” In Proceedings of the 4th International Conference on Speech Prosody, Campinas Brazil, 6-9

318

References

May 2008, edited by Plinio A. Barbosa, Sandra Madureira, and César Reis, 739-42. Campinas, Brazil: Editora RG/CNPq, 2008. Mori, Laura. Fonetica dell’italiano L2. Un’indagine sperimentale sulla variazione nell’interlingua dei marocchini. Roma: Carocci, 2007. Moulines, Eric, and Michel Charpentier. “Pitch-synchronous waveform processing techniques for text-to-speech synthesis using diphones.” Speech Communication 9 (1990): 453-67. Muljaþiü, Zࡅ arko. Fonologia della lingua italiana. Bologna: Il Mulino, 1972. Nakatani, Christine H., and Julia Hirschberg. “A corpus-based study of repair cues in spontaneous speech.” Journal of acoustical society of America 95 (2009): 1603-16. Nesterenko, Irina. “Prosodic annotation in spontaneous Russian: perceptual annotation and automatic classification, speech prosody.” In Proceedings of the 3rd International Conference on Speech Prosody, Dresden, Germany, 2-5 May 2006, edited by Hoffmann Rüdiger and Hansiörg Mixdorff. TUD PRESS 2006, available to http://aune.lpl.univaix.fr/~sprosig/sp2006/contents/papers/PS529_0110.pdf. Nickerson, R. “Characteristics of the speech of deaf persons.” The Volta Review 77 (1975): 342–62. Nooteboom, Sieb G. “Speaking and unspeaking: detection and correction of phonological and lexical errors in spontaneous speech.” In Errors in linguistic performance, edited by Victoria Hoffmann Rüdiger, and Hansiörg Mixdorff, 287-305. New York: Academic Press, 1980. Nozawa, Takeshi, and Sang Yee Cheon. “How listeners’ L1 affects perceived similarity of American English, Japanese and Korean vowels.” Paper presented at ICPhS XVII, Hong Kong, August 17-21, 2011. O’Connor, Joseph Desmond, and Gordon Arnold. Intonation of Colloquial English. London: Longman, 1973. Okalidou, Areti, and Katherine S. Harris. “A comparison of intergestural patterns in deaf and hearing adult speakers: Implications from an acoustic analysis of disyllables.” Journal of the Acoustical Society of America 106 (1999): 394–410. Pallotti, Gabriele. “Variations situationelles dans la construction des énoncés en L2: le cas des autorépétitions.” AILE - Acquisition et interaction en langue étrangère 22 (2005): 101-30. Paoloni, Andrea. “La voce come elemento di identificazione della persona.” In La voce come bene culturale, edited by Amedeo De Dominicis, 125-39. Roma: Carocci, 2002.

Prosodic and Rhythmic Aspects of L2 Acquisition: The Case of Italian

319

Péan, Vincent, Sheila Williams, and Maxine Eskenazi. “The design and recording of icy, a corpus for the study of intraspeaker variability and the characterisation of speaking styles.” In EUROSPEECH 1993, 62730. Berlin: European Speech Communication Association, 1993. Pérez i Parent, Montserrat. “The production of requests by Catalan learners of English: Situational and proficiency level effects.” ATLANTIS 2 (2002): 147–68. Peters, Jörg. “A bitonal lexical pitch accent in the Limburgian dialect of Borgloon.” In Tones and Tunes. Volume 1: Typological Studies in Word and Sentence Prosody, edited by Tomas Riad, and Carlos Gussenhoven, 167-98. Berlin: Mouton de Gruyter, 2007. ʊ. “Tone and intonation in the dialect of Hasselt.” Linguistics 46/5 (2008): 983-1018. Pettorino, Massimo. “I cambiamenti della lingua italiana.” In La voce come bene culturale, edited by Amedeo De Dominicis, 141-57. Roma: Carocci, 2002. ʊ. “La velocità di articolazione.” In Costituzione, gestione e restauro di corpora vocali, edited by Amedeo De Dominicis, Laura Mori, and Marianna Stefani, 227-32. Roma: Esagrafica, 2003. ʊ. “Inglese italiano e giapponese: analisi dei correlati acustici delle emozioni nel parlato cinematografico.” In Comunicazione parlata e manifestazione delle emozioni. Atti del 1° Convegno Nazionale GSCP Gruppo di Studio della Comunicazione Parlata. Padova 30 Novembre - 1 Dicembre 2004, edited by Emanuela Magno Caldognetto, Federica Cavicchio, and Piero Cosi, 45-57. Napoli: Liguori, 2004. Pettorino, Massimo, and Antonella Giannini. “Analisi delle disfluenze e del ritmo di un dialogo romano.” In Italiano parlato. Analisi di un dialogo, edited by Federico Albano Leoni, and Rosa Giordano, 89-104. Napoli: Liguori, 2005. ʊ, “Il parlato dei mass media.” In La Comunicazione Parlata 3, edited by Massimo Pettorino, Antonella Giannini, and Francesca M. Dovetto, 71-83. Napoli: OPAR, Università degli Studi di Napoli “L’Orientale”, 2010. Pettorino, Massimo, Antonella Giannini and Francesca Dovetto, eds. La comunicazione parlata 3. Napoli: Università degli Studi di Napoli “L’Orientale”, 2010. Petty, John, Richard E., and John T. Cacioppo. “Issue Involvement as a Moderator of the Effects on Attitude of Advertising on Content and Context.” Advances in Consumer Research 8 (1981): 20–4.

320

References

Pickett, Emily R., Sheila E. Blumstein, and Martha W. Burton. “Effects of Speaking Rate on the Singleton-Geminate Consonant Contrast in Italian.” Phonetica 56 (1999): 135-57. Pierrehumbert, Janet B. “The phonology and phonetics of English intonation.” PhD diss., Massachusetts Institute of Technology, 1980. Bloomington: Indiana University Liberian Collections, 1988. Plauché, Madeleine C., and Elizabeth. E. Shriberg. “Data-driven subclassification of disfluent repetitions based on prosodic features.” In Proceedings of the International Congress of Phonetic Sciences, San Francisco, USA, Berkeley, University of California, edited by Ohala, John J., Yoko Hasegawa, Manjari Ohala, Daniel Granville, and Ashlee C. Bailey, vol. 2, 1513-16, 1999. Poggi, Isabella, and Emanuela Magno Caldognetto. “Il Parlato emotivo. Aspetti cognitivi, linguistici e fonetici.” In Atti del Convegno “Il parlato emotivo”. Napoli, 12-15 febbraio 2003. Napoli: D’Auria Cd Rom, 2003. Port, Robert F. “Linguistic timing factors in combination.” Journal of the Acoustical Society of America 69 (1981): 262-74. Queen, Robin M. “Bilingual intonation patterns: Evidence of language change from Turkish-German bilingual children.” Language in Society 30 (2001): 55–80. Ramus, Franck, Emmanuel Dupoux, and Jacques Mehler. “The psychological reality of rhythm classes: perceptual studies.” In Proceedings of the 15th International Congress of Phonetic Sciences, Barcelona, Spain 3-9 August 2003, edited by Solé Maria Josep, Recasens Daniel and Joaquín Romero, 337-42. Barcelona: Universitat Autonòma de Barcelona, 2003. Rasier, Laurent. “Prosodie en vreemdetaalverwerving. Accentdistributie in het Frans en het Nederlands als vreemde taal.” PhD diss., Université catholique de Louvain, 2006. Rasier, Laurent, and Philippe Hiligsmann. “Prosodic transfer. Theoretical and Methodological Issues.” Nouveaux cahiers de linguistique Française 28 (2007): 41-66. Rastelli, Stefano, ed. Italiano di cinesi, italiano per cinesi. Dalla prospettiva della didattica acquisizionale. Perugia: Guerra, 2010. Rauniomaa, Mirka. “Recovery through repetition. Returning to prior talk and taking a stance in American- English and Finnish conversations.” Academic dissertation, Faculty of Humanities of the University of Oulu, 60. Oulu: Oulu University Press, 2008. Ridouane, Rachid. “Geminates at the junction of phonetics and phonology.” In Papers in Laboratory Phonology 10, edited by Cécile

Prosodic and Rhythmic Aspects of L2 Acquisition: The Case of Italian

321

Fougeron, Barbara Kühnert, Mariapaola D’Imperio, and Nathalie Vallée, 61-80. Berlin / New York: De Gruyter, 2010. Romano, Antonio, and Grazia Maria Interlandi. “Variabilità geo-socioprosodica: dati linguistici e statistici.” Géolinguistique 3 (2005): 25980. Hors série: Projet AMPER - Atlas Multimédia Prosodique de l’Espace Roman, 2005. Romito, Luciano, Renata Savy, Andrea Tarasi, and Rosita Lio. “Analisi percettiva, musicale e automatica dell’italiano L1 e L2.” Paper presented at the annual conference for the A.I.S.V., Rome, January 2527, 2012, in press. Romito, Luciano, Andrea Tarasi, and Rosita Lio. “Italian Index: rhytmical-prosodic analysis of Italian L2 produced by Albanian, Chinese, Polish and Romanian speakers.” Paper presented at the biennal conference for the V C.F.E., Càceres, Extremadura, October 25-28, 2011, in press. Romito, Luciano, and John Trumper. “Problemi teorici e sperimentali posti dall’isocronia.” In Quaderni del Dipartimento di Linguistica dell’Università della Calabria, edited by Cattedra di Linguistica, 88115. Cosenza: University of Calabria, 1993. Rose, Kenneth R. “An exploratory cross-sectional study of interlanguage pragmatic development.” Studies in Second Language Acquisition 22, (2000): 27-67. Rossi, Fabio. Tratti pragmatici e prosodici della dislocazione a destra nel parlato spontaneo. LABLITA, 7. Firenze: Università di Firenze, 2000. Russo, Domenico. “I ‘colori’ della semantica regionale. Indagine sui termini del lessico alimentario abruzzese.” Plurilinguismo. Contatti di lingue e culture 8 (2001): 111-26. ʊ. “Per lo studio dei lessici degli italiani regionali. I Materiali per un lessico alimentario abruzzese.” In Studi e ricerche di terminologia alimentaria, edited by Carlo Consani, 27-51 and 91-410. Alessandria: Edizioni dell’Orso, 2001. Russo, Domenico. “Sulla prosodia dell’italiano regionale abruzzese (IRA).” In L’italiano parlato, edited by Federico Albano Leoni, Francesco Cutugno, Massimo Pettorino, and Renata Savy, CD-ROM E06. Napoli: D’Auria, 2004. Sabatini, Francesco. “L’italiano dell’uso medio: una realtà tra le varietà linguistiche italiane.” In Gesprochenes Italienisch in Geschichte und Gegenwart, edited by Gunther Holtus, and Edgar Radtke, 154-83. Tübingen: Narr, 1985. Sacks, Harvey. “Notes on Methodology.” In Structures of social action: studies in conversation analysis, edited by John Maxwell Atkinson,

322

References

and John Heritage, 21-7. Cambridge: Cambridge University Press, 1984. ʊ. Lectures on conversation, Vol. I. and II. Oxford: Blackwell, 1992. Sacks, Harvey, Emanuel A. Schegloff, and Gail Jefferson. “A simplest systematics for the organisation of turn taking for conversation.” Language 50 (1974): 696-735. Sala Gallini, Mario. “Lo statuto del clitico nella dislocazione a destra: pronome vero o marca flessionale?” Archivio glottologico italiano. LXXXI/1 (1996): 76-94. Salvati, Luisa. “Il debating interculturale. L’interfaccia prosodicoargomentativa nel parlato argomentativo in Italiano L1/L2.” PhD diss., University of Naples “L’Orientale”, 2012. Salza, Pier Luigi, Claudia Barolo, and Beata Dobrzynska. “Etichettatura fonetica e TTS.” Paper presented at the IV National Congress AISV, Arcavacata di Rende, Italy, December 3-5, 2007. Salza, Pier Luigi, Enzo Foti, Luciano Nebbia, and Mario Oreglia. “MOS and Pair Comparison Combined Methods for Quality Evaluation of Text-to-Speech Systems.” Acta Acustica united with Acustica 82 (1996): 650-56. Saussure, Ferdinand de. Course in General Linguistics. Translated by Wade Baskin. New York: Philosophical Library, 1959. Savastano, Elvira, Antonella Giannini, and Massimo Pettorino. “Aspetti prosodici del parlato dei politici.” In Atti del XXIII Congresso Nazionale dell’AIA, 171-76. Bologna: Editore Cocchi, 1995. Savino, Michelina. “Non-finality and Pre-finality in Bari Italian Intonation: a Preliminary Account.” In Eurospeech 2001 Scandinavia: 7th European Conference on Speech Communications and Technology, 3-7 September 2001, edited by Dalsgaard Paul, Lindberg Børge, 93942. Aalborg, Denmark: Kommunik Grafiske Løsninger A/S, 2001. ʊ. “Intonational Cues to Discourse Structure in a Variety of Italian.” In Regional Variation in Intonation, edited by Peter Gilles, and Jörg Peters, 145-59. Tübingen: Niemeyer, 2004. Savino, Michelina, Martine Grice, Barbara Gili Fivela, and Giovanna Marotta. “Intonational cues to discourse structure in Bari and Pisa Italian: perceptual evidence.” In Proceedings of the 3rd International Conference on Speech Prosody, Dresden, Germany, 2-5 May 2006, edited by Hoffmann Rüdiger, and Hansiörg Mixdorff, 114-17. Dresden: TUD PRESS, 2006. Savy, Renata. “Pr.A.T.I.D: a coding scheme for pragmatic annotation of dialogues.” In Proceedings of LREC 2010, edited by Núria Bel, Béatrice Daile, and Andrejs Vasiljevs, 2141-48. Malta, 2010.

Prosodic and Rhythmic Aspects of L2 Acquisition: The Case of Italian

323

Savy, Renata, and Marina Castagneto. “Funzioni comunicative e categorie d’analisi pragmatica: dal testo dialogico allo schema xml e viceversa.” In Linguistica e modelli tecnologici di ricerca, Atti del XL Congresso SLI, edited by Giacomo Ferrari, 569-79. Rome: Bulzoni, 2006. Savy, Renata, and Francesco Cutugno. “Su alcune correlazioni tra riduzioni segmentali e tratti prosodici nel parlato spontaneo: il ruolo del fattore tempo.” In Atti del XXVII Convegno Nazionale dell’Associazione Italiana di Acustica, Genova, 26-28 Maggio 1999, 183-87. ʊ. “CLIPS. Diatopic, diamesic and diaphasic variations in spoken Italian.” In On-line Proceedings 5th Corpus Linguistics Conference, edited by Michaela Mahlberg, Victorina González Díaz, and Catherine Smith, University of Liverpool, UK, 2009, accessed June 2012, http://ucrel.lancs.ac.uk/ publications/cl2009/ article 213. Schegloff, Emanuel A. “Third Turn Repair.” In Towards a Social Science of Language. Social interaction and discourse structures. Papers in honour of William Labov, edited by Gregory R. Guy, Crawford Feagin, Deborah Schiffrin, and John Baugh, Vol. 2, 31-40. Amsterdam: John Benjamins, 1997. ʊ. “On dispensability.” Research on language and social interaction 37/2 (2004): 95–149. Schegloff, Emanuel A., and Harvey Sacks. “Opening up closings.” Semiotica 8 (1973): 289-327. Scherer, Klaus R. “Vocal communication of emotion: A review of research paradigms.” Speech Communication 40 (2003): 227-56. Scherer, Klaus R., Tom Johnstone, and Gudrun Klasmeyer. “Vocal expression of emotion.” In Handbook of the Affective Sciences, edited by Richard J. Davidson, Klaus R. Scherer, and Hill Goldsmith, 433-56. New York / Oxford: Oxford University Press, 2003. Schilperoord, Joost. It’s about time. Temporal aspects of Cognitive Processes in Text. Amsterdam-Atlanta, GA: Rodpi (Utrecht Studies in language and communication), 1996. Schröder, Marc. “Speech and Emotion Research: An overview of research frameworks and a dimensional approach to emotional speech synthesis.” PhD thesis, Saarland University, 2008. Schuetze-Coburn, Stephen, and Elizabeth G. Weber. “Units of intonation in discourse: a comparison of acoustic and auditory analyses.” Language and Speech 34, (1991): 207-37. Searle, John. Speech Acts. Cambridge: Cambridge University Press, 1969. Selinker, Larry. “Interlanguage.” International Review of Applied Linguistics in Language Teaching X/3 (1972): 209-31.

324

References

Selkirk, Elisabeth. “The prosodic structure of function words.” In Papers in Optimality Theory: University of Massachusetts Occasional Papers Volume 18, edited by Jill N. Beckman, Laura Walsh Dickey, and Suzanne Urbanczyk, 439–69. Amherst: GLSA, UMass–Amherst, 1995. Shields, Lynne W., and David A. Balota. “Repetition and associative context effects in speech production.” Language and speech 31 (1991): 47-55. Shriberg Elizabeth E. “Phonetic consequences of speech disfluency.” In Proceedings of the International Congress of Phonetic Sciences, edited by Ohala, John J., Yoko Hasegawa, Manjari Ohala, Daniel Granville, and Ashlee C. Bailey, vol. 1, 619-22. San Francisco, USA, Berkeley: University of California, 1999. ʊ. “To ‘errrr’ is human: ecology and acoustics of speech disfluencies.” Journal of International Phonetic Association 31 (2001): 153-69. Shriberg, Lawrence D., Joan Kwiatkowski, and Carmen Rasmussen. Prosody–Voice Screening Profile (PVSP). Tuscon, AZ: Communication Skill Builders, 1990. Silva, Gláucia, and Denise Santos. “Framing participation through repetition: the case of a Portuguese learner in different settings.” Portuguese Language Journal 1 (2006): 1-24. Simone, Raffaele. “Une interprétation diachronique de la ‘dislocation à droite’ dans les langues romanes.” Langue Française 15 (1997): 48-61. Smith, Caroline L. “The temporal organization of vowels and consonants.” Unpublished PhD diss., Yale University, 1992. Smith, Stephen M., and David R. Shaffer. “Celerity and Cajolery: Rapid Speech May Promote or Inhibit Persuasion through its Impact on Message Elaboration.” Personality and Social Psychology Bulletin 17 (1991): 663-69. ʊ. “Speed of Speech and Persuasion: Evidence for Multiple Effects.” Personality and Social Psychology Bulletin 21(1996): 1051-60. Sobrero, A. Alberto. “Italienisch: Regionale Varianten.” In Lexicon der Romanistischen Linguistik (LRL), edited by Günther Holtus, Michael Metzeltin and Christian Schmitt, 732-48. Tübingen: Niemeyer, 1988. Sorianello, Patrizia. “Modelli intonativi dell’interrogazione in una varietà di italiano meridionale.” Rivista Italiana di Dialettologia XXV (2001): 85-108. ʊ. Prosodia. Roma: Carocci, 2006. ʊ. “Per una definizione fonetica dei confini prosodici.” In La comunicazione parlata/Spoken Communication, edited by Massimo

Prosodic and Rhythmic Aspects of L2 Acquisition: The Case of Italian

325

Pettorino, Antonella Giannini, and Renata Savy, 310-30. Napoli: Liguori, 2008. Sorjonen, Marja-Leena. “Vuoronalkuiset konnektorit: mutta [Turn-initial connectors: mutta ‘but’].” In Kieli 4: Suomalaisen keskustelun keinoja I [Characteristics of Finnish conversation I], edited by Auli Hakulinen, 162–76. Helsinki: Department of Finnish, University of Helsinki, 1989. Sornicola, Rosanna. Sul parlato. Bologna: Il Mulino, 1981. Spaai, Gerard, and Dik Hermes. “A visual display for the teaching of intonation.” CALICO Journal 10 (1993): 19-30. Sperber, Dan, and Deirdre Wilson. Relevance: Communication and Cognition. Oxford: Blackwell, 1995. Stathopoulos, Elaine, Judit Duchan, Rae M. Sonnenmeier, and Nancy V. Bruce. “Intonation and pausing in deaf speech.” Folia Phoniatrica 38 (1986): 1–12. Stivers, Tanya. “ No no no and other types of multiple sayings in social interaction.” Human Communication Research 30/2 (2004): 260–93. Stuart, Stewart and Lynn Pearson. “Development of communication strategies among foreign language learners.” Mountain interstate foreign language review 5 (1995): 122-27. Stutterheim, Christian von. “Global principles of information organisation in texts of L2 speakers.” In Studi Italiani di Linguistica Teorica e Applicata, Strutture testuali e principi di organizzazione dell’informazione nell’apprendimento linguistico, edited by Marina Chini, and Anna Giacalone Ramat, 89-110. Roma: Pacini Editore, 1998. Subtelny, Joanne, Robert Whitehead, and Nicholas A. Orlando. “Description and evaluation of an instructional program to improve speech and voice diagnosis of the hearing impaired.” The Volta Review 82 (1980): 85–95. Swerts, Mark. “Prosodic features at discourse boundaries of different strength.” Journal of Acoustical Society of America 101 (1997): 51421. Swerts, Mark, Don G. Bouwhuis, and René Collier. “Melodic cues to the perceived ‘finality’ of utterances.” Journal of Acoustical Society of America 96 (1994): 2064-75. Swerts, Mark, René Collier, and Jacques M. B. Terken. “Prosodic predictors of discourse finality in spontaneous monologue.” Speech Communication 15 (1994): 79-90. Swerts, Mark, and Ronald Geluykens. “The prosody of information units in spontaneous monologue.” Phonetica 50 (1993): 189-96.

326

References

ʊ.“Prosody as a marker of information flow in spoken discourse.” Language and Speech 37/1 (1994): 21-43. Swerts, Mark, and Sabine Zerbian. “Prosodic transfer in Black South African English.” In Proceedings of the 5th International Conference on Speech Prosody Chicago, Illinois 11-14 May, 2010. Illinois 2010, University of Illinois Press, available to http://speechprosody2010. illinois.edu/program.php /papers/100198.pdf. Tannen, Deborah. Talking voices: repetition, dialogue and Imagery in Conversational Discourse. 2nd edition. New York: Cambridge University Press, 2007. Telmon, Tullio. Guida allo studio degli italiani regionali. Alessandria: Edizioni dell’Orso, 1990. ʊ. “Varietà regionali.” In Introduzione all’italiano contemporaneo. La variazione e gli usi, edited by Alberto Sobrero, 93-149. Roma: Laterza, 1993. ʊ. “Gli italiani regionali contemporanei.” In Storia della lingua italiana, Vol. III, Le altre lingue, edited by Luca Serianni, and Pietro Trifone, 597-626. Torino: Einaudi, 1994. Tepperman, Joseph. “Testing suprasegmental English through Parroting.” Paper presented at the biennal meeting for the Speech Prosody, Chicago, Illinois, May 11-14, 2010. Tepperman, Joseph, and Emily Nava. “Long-distance rhythmic dependencies and their application to automatic language identification.” Paper presented at the annual conference for the Interspeech, Florence, Toscana, August 28-31, 2011. Tomlin, Russell. “Functionalism in Second Language Acquisition.” Studies in Second Language Acquisition XII (1990): 155-77. Traum, David. Coding Schemes for Spoken Dialogue Structure, Unpublished manuscript, 1996. Trofimovich, Pavel, and Wendy Baker. “Learning second-language suprasegmentals: Effect of L2 experience on prosody and fluency characteristics of L2 speech.” Studies in Second Language Acquisition 28 (2006): 1-30. ʊ. “Learning prosody and fluency characteristics of second language speech: The effect of experience on child learners’ acquisition of five suprasegmentals.” Applied Psycholinguistics 28 (2007): 251–76. Trosborg, Anna. Interlanguage Pragmatics: Requests, Complaints and Apologies. Berlin / New York: Mouton Gruyter, 1995. Trouvain, Jürgen, and Ulrike Gut, eds. Non-Native Prosody. Phonetic description and teaching practice. Berlin: Mouton de Gruyter, 2007.

Prosodic and Rhythmic Aspects of L2 Acquisition: The Case of Italian

327

Tsalikis, John, Oscar W. DeShields, Jr, and Michael S. LaTour. “The role of accent on the credibility and effectiveness of the salesperson.” Journal of Personal Selling and Sales Management 11 (1991): 31-41. Tyler, Michael D. and Anne Cutler. “Cross-language differences in cue use for speech segmentation.” Journal of Acoustical Society of America 126 (2009): 367-76. Ueyama, Motoko. “The phonology and phonetics of second language intonation: the case of ‘Japanese English’.” In Proceedings of Eurospeech ’97, edited by George Kokkinakis, Nikos Fakotakis, and Evangelos Dermatas, 2411-14. Rhodes: University of Patras, 1997. Ueyama, Motoko, and Sun-Ah Jun. “Focus realization in Japanese English and Korean English intonation.” In Japanese/Korean Linguistics, edited by Noriko Akatsuka, Hoji Hajime, and Shoichi Iwasaki, 7: 62945. Stanford: Center for the Study of Language and Information, 1998. Ullakonoja, Riikka. “Pausing as an indicator of fluency in the Russian of Finnish learners.” In Proceedings of the 4th International Conference on Speech Prosody, Campinas Brazil, 6-9 May 2008, edited by Plinio A. Barbosa, Sandra Madureira, and César Reis, 339-42. Campinas, Brazil 2008: Editora RG/CNPq. Vaissière, Jacqueline. “Phonetic explanations for cross-linguistic prosodic similarities.” Phonetica 52 (1995): 123-30. van den Berg, Rob, Carlos Gussenhoven, and Toni Rietveld. “Downstep in Dutch: Implications for a model.” In Papers in laboratory phonology II: Gesture, Segment, Prosody, edited by Gerard J. Docherty, and D. Robert Ladd, 335-59. Cambridge: Cambridge University Press, 1992. van Heuven, Vincent J., and Judith Haan. “Phonetic correlates of statement versus question intonation in Dutch.” In Intonation. Analysis, Modelling and Technology, edited by Antonis Botinis, 11943. Dordrecht: Kluwer Academic Publishers, 2000. Vardanian, Rose Marie. “Teaching English through oscilloscope displays.” Language Learning 3-4 (1964): 109-11. Vedovelli, Massimo. L’italiano degli stranieri. Roma: Carocci, 2002. ʊ. “Lingue immigrate nel Mediterraneo e nuove modalità di rilevazione sociolinguistica.” In Mediterraneo plurilingue. Atti del convegno di studi, Genova, 13-15 maggio 2004, edited by Vincenzo Orioles, and Fiorenzo Toso, 363-83. Udine: Università degli Studi, Centro di studi sul plurilinguismo, 2009. Vedovelli, Massimo, Stefania Massara, and Anna Giacalone Ramat. Lingue e culture in contatto. L’italiano come L2 per gli arabofoni. Milano: Franco Angeli, 2001.

328

References

Verhoeven, Jo. “Belgian Standard Dutch.” Journal of the International Phonetic Association 35 (2005): 243-47. Vitagliano, Ilaria, Antonella Giannini, and Massimo Pettorino. “Le emozioni nel tempo.” In Fenomeni di intensità nell’italiano parlato, edited by Barbara Gili Fivela, and Carla Bazzanella. Firenze: Franco Cesati, 2008. Voghera, Miriam. Sintassi e intonazione dell’italiano parlato. Bologna: Il Mulino, 1992. ʊ. “Teorie linguistiche e dati di parlato.” In Dati empirici e teorie linguistiche. Atti del XXXIII Congresso della SLI (Napoli, 28– 30/10/1999), edited by Federico Albano Leoni, Eleonora Stenta Krosbrakken, Rosanna Sornicola, and Carolina Stromboli, 75–95. Roma: Bulzoni, 2001. Wagner, Agnieszka. “Acoustic cues for automatic determination of phrasing.” In Proceedings of the 5th International Conference on Speech Prosody Chicago, Illinois, 11-14 May, 2010, Illinois 2010, University of Illinois Press, available to http://speechprosody2010.illinois.edu/program.php/papers/100196.pdf. Wells, John C. “SAMPA computer readable phonetic alphabet.” Last revised October 25, 2005, http://www.phon.ucl.ac.uk/home/sampa. Whipple, Thomas W., and Mary K. McManamon. “Implications of using male and female voices in commercials: An exploratory study.” Journal of Advertising 31 (2002): 79-91. Wiberg, Eva. Il riferimento temporale nel dialogo. Un confronto tra giovani bilingui italo-svedesi e giovani monolingui romani. Études romanes de Lund 58. Lund: University Press, 1997. Wichmann, Anne, and Blakemore Diane. “The prosody-pragmatics interface.” Journal of Pragmatics 38 (2006): 1537-41. Willis, Jane. A framework for task-based learning. Harlow: U. K. Longman Addison-Wesley, 1996. Wong, Jean. “Repetition in conversation: A look at ‘first and second sayings’.” Research on language and social interaction 33/4 (2000): 407–24. Xianonan, Shen Susan. “Ability of Learning Prosody of an Intonational Language by Speakers of a Tonal Language: Chinese Speakers Learning French Prosody.” IRAL 2 (1990): 119-34. Yang, Li-Chiung. “Duration and pauses as phrasal and boundary marking indicators in speech.” In Proceedings of the 15th International Congress of Phonetic Sciences, Barcelona, Spain 3-9 August 2003, edited by Solé Maria Josep, Recasens Daniel, and Joaquín Romero, 1791-94. Barcelona: Universitat Autonòma de Barcelona, 2003.

Prosodic and Rhythmic Aspects of L2 Acquisition: The Case of Italian

329

Zhang, Xiaoling. “Echoing in real-life English conversation.” PALA: The Poetics and Linguistics Association, Occasional paper n. 9, 1998, 1-16. Zmarich, Claudio, Barbara Gili Fivela, Pascal Perrier, Christophe Savariaux, and Graziano Tisato. “Consonanti scempie e geminate in Italiano: studio acustico e cinematico dell’articolazione linguale e bilabiale.” In Atti del III Convegno Nazionale dell’Associazione Italiana di Scienze della Voce (AISV), edited by Veronica Giordani, Valentina Bruseghini, and Piero Cosi, 151-63. Torriana: EDK Editore, 2007. ʊ. “Speech timing organization for the phonological length contrast in Italian consonants.” In Proceedings of Interspeech 2011, 28-31 August, 2011, 401-04. Firenze, Italy, 2011.

CONTRIBUTORS IOLANDA ALFANO is a PhD student in Spanish Philology at the Autonomous University of Barcelona, Spain. Her research interests, and more recent publications, include the interface between pragmatics and prosody, second language acquisition and contrastive studies between Italian and Spanish. [email protected] JENS BAELE is a young researcher at Ghent University, Belgium. He specializes in Italian Linguistics and currently works on the acquisition of foreign language intonation. [email protected] VALERIA CARUSO is a Post-Doctoral researcher at the University of Naples “L’Orientale”, Italy. Her research interests include the relationship between wordplay and poetry, translation, electronic lexicography, and second language acquisition and teaching deaf immigrants. [email protected] CHIARA CELATA is a Post-Doctoral researcher in Linguistics and Phonology at Scuola Normale Superiore of Pisa, Italy. Her recent works focus on experimental phonetics and phonology, the acquisition of L2 Italian phonology and on the relationship between L2 phonological acquisition and historical change. [email protected] CLAUDIA CROCCO is Professor of Italian Linguistics at Ghent University, Belgium. Her research is centred on spoken language and includes intonational phonetics and phonology, the relation between prosody and pragmatics and the acquisition of the intonation of Italian as a foreign language. [email protected] LIDIA COSTAMAGNA is Professor of Phonology and Psycholinguistics at University for Foreigners of Perugia, Italy. She is the author of many articles and books on L2 Italian phonological acquisition. [email protected]

332

Contributors

ALBERT DE GROOT joined Loquendo, now part of Nuance Communication, in 2003, as Casting Director and Producer, specializing in directing Voice Talents for TTS-system speech database recording. He is also a translator, theatre director, visual artist and musician. [email protected] ANNA DE MARCO is researcher at the University of Calabria, Italy. She teaches courses in General and Applied Linguistics. She has published extensively on L1 Acquisition, with a focus on morphology, and on language teaching and learning. Her research interests include pragmatics and sociopragmatics. [email protected] ANNA DE MEO is Professor at the University of Naples “L’Orientale”, Italy. She teaches Multiculturality and Language Acquisition and Specialized Translation. Her main fields of interest are L2 acquisition (with special reference to Italian), L2 speech perception and production, interlanguage pragmatics, and translation. [email protected] MARTA MAFFIA is a PhD student in Linguistics at the University of Naples “L’Orientale”, Italy, and her main fields of interest are pragmatics, L2 prosody, intercultural communication and language acquisition. [email protected] PHILIPPE MARTIN has been teaching at the University of Toronto and Aix-en-Provence, and is head of the Linguistic Department at the University Paris Diderot. He has published numerous papers on intonation phonology and phonetics. His last two books were Phonétique Acoustique and Intonation du Français. [email protected] LAURA MORI teaches Sociolinguistics, Contact Linguistics and Text Linguistics and Pragmatics at the LUSPIO University of Rome, Italy. Her main research areas are phonetic variation in L2 Italian by Moroccan migrants, contact linguistics, translation-applied linguistics and variational sociolinguistics. [email protected]

Prosodic and Rhythmic Aspects of L2 Acquisition: The Case of Italian

333

ELISA PELLEGRINO is a Post-Doctoral researcher in Experimental Phonetics at University of Naples “L’Orientale”, Italy. Her research interests deal with foreign-accented Italian, L2 speaker recognition based on acoustic cues and with rhythmic and prosodic aspects of L2 acquisition in deaf learners. [email protected] MASSIMO PETTORINO is Professor of Experimental Phonetics at the University of Naples “L’Orientale”, Italy, and Director of the GSCP, a special interest group on spoken communication of the Italian Linguistic Society (SLI). He works in the field of acoustic and prosodic analysis of speech. [email protected] MILENA ROMANO is a PhD student in Modern Philology - Curriculum editing at the University of Catania, Italy. She is working on the use of rotogravure as a means of language standardization in the WWII post-war period. She is a member of a research team on television language, coordinated by Prof. Gabriella Alfieri. [email protected] LUCIANO ROMITO is Professor of Linguistics and Experimental Phonetics at the University of Calabria, Italy. He is Director of Laboratory of Phonetics, President of Italian Association of Voice Science and National Coordinator of the Forensic Phonetics Group. He works also on dialectology and sociolinguistics. [email protected] DOMENICO RUSSO is Professor of Italian Linguistics and Philosophy of Language at “G. D’Annunzio” University of Chieti-Pescara. He focuses on epistemology of linguistics, semantics, lexicography and prosody. He wrote Accentuario del lessico italiano and Sillabario del lessico italiano. [email protected] LUISA SALVATI is a Post-Doctoral researcher in L2 acquisition and teaching at the University for Foreigners, Siena, Italy. She teaches Multiculturality and Language Acquisition at the University of Naples “L’Orientale”. Her research interests also include pragma-phonetics and intercultural communication. [email protected]

334

Contributors

PIER LUIGI SALZA has been working on TTS technology as phonetician since 1982, formerly at CSELT and then at Loquendo, now part of Nuance Communication. His interests are concatenative unit design, phonetic transcription rule development, prosody and subjective evaluation and testing of TTS. [email protected] RENATA SAVY is Professor of Linguistics and Applied Linguistics at the University of Salerno, Italy. Her research interests focus on spoken corpora, and include phonetics, prosody and pragmatics, and the acquisition of Italian as a foreign language. [email protected] PATRIZIA SORIANELLO is Professor of Linguistics at the University of Bari, Italy. Her recent interests include the acquisition of L2 Italian phonology, second language speech perception, the intonation of Southern Italian varieties, and speech pathology. [email protected] ANDREA TARASI is a PhD student in Linguistics at the University of Salerno, Italy, with a thesis on the rhythmic aspects of some regional varieties of Italian. He is also interested in rhythmic and prosodic analysis of L2 Italian produced by Albanian, Chinese, Polish and Romanian speakers. [email protected] MARILISA VITALE is a PhD student in Linguistics at the University of Naples “L’Orientale”, Italy. Her main fields of interest are rhythmicprosodic aspects of L2 Italian, foreign accent correlates and credibility, speech synthesis, teaching L2 Italian prosody. [email protected] ENRICO ZOVATO is R&D Engineer at Loquendo, now part of Nuance Communication, and he is member of W3C and of the ISCA. He works on speech data acquisition, human-machine interaction, statistical parametric synthesis, acoustic parameter prediction and generation for expressive speech synthesis. [email protected]