The Intonational Phonology of Swabian and Upper Saxon 9783484305151, 9783110932218

The study employs an autosegmental-metrical model of intonation to propose an intonational grammar of Swabian and Upper

282 76 8MB

German Pages 204 Year 2007

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
Acknowledgments
Contents
1. Introduction
2. The empirical basis
3. Modeling dialect intonation
4. Swabian German Intonation Patterns
5. Upper Saxon German Intonation Patterns
6. Realisational differences between Swabian and Upper Saxon German
7. A note on intonational meaning
8. Summary and Conclusion
A Map Tasks
Bibliography
Recommend Papers

The Intonational Phonology of Swabian and Upper Saxon
 9783484305151, 9783110932218

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

Linguistische Arbeiten

515

Herausgegeben von Klaus von Heusinger, Gereon Müller, Ingo Plag, Beatrice Primus, Elisabeth Stark und Richard Wiese

Frank Kügler

The Intonational Phonology of Swabian and Upper Saxon

Max Niemeyer Verlag Tübingen 2007

Für Astrid und Liv

Bibliografische Information der Deutschen Nationalbibliothek Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über http://dnb.ddb.de abrufbar. ISBN 978-3-484-30515-1

ISSN 0344-6727

© Max Niemeyer Verlag, Tübingen 2007 Ein Imprint der Walter de Gruyter GmbH & Co. KG http.V/www.niemeyer. de Das Werk einschließlich aller seiner Teile ist urheberrechtlich geschützt. Jede Verwertung außerhalb der engen Grenzen des Urheberrechtsgesetzes ist ohne Zustimmung des Verlages unzulässig und strafbar. Das gilt insbesondere für Vervielfältigungen, Übersetzungen, Mikroverfilmungen und die Einspeicherung und Verarbeitung in elektronischen Systemen. Printed in Germany. Gedruckt auf alterungsbeständigem Papier. Druck und Einband: AZ Druck und Datentechnik GmbH, Kempten

Acknowledgments

Das vorliegende Buch ist eine überarbeitete Version meiner an der Universität Potsdam im Jahre 2005 eingereichten Dissertation mit dem Titel "Swabian and Upper Saxon Intonational Patterns". An dieser Stelle möchte ich mich bedanken bei allen, die in der unterschiedlichsten Weise zur Entstehung dieser Arbeit beigetragen haben. Mein besonderer Dank geht an meine beiden Betreuer, Caroline Fery und Jörg Mayer. Beide haben in unermüdlicher Diskussionsbereitschaft meine Gedanken zur Intonation reifen lassen. Caroline Fery hat darüber hinaus mit einem Stipendium im Tübinger Graduiertenkolleg "Integriertes Linguistik Studium" und einer Mitarbeiterstelle an der Universität Potsdam die finanzielle Grundlage dieser Arbeit geschaffen. Weiterhin danke ich Peter Auer, der mit hilfreichen Kommentaren und Diskussionen seit der Initiative dieser Dissertation beteiligt war. Schließlich bin ich Richard Wiese zu großem Dank verpflichtet, der als Herausgeber der Serie "Linguistische Arbeiten" viele wertvolle und hilfreiche Hinweise zur Verbesserung des Manuskripts dieser Arbeit gegeben hat. Ohne die Bereitschaft der Dialektsprecher aus Stuttgart und Leipzig hätte diese Arbeit so nicht entstehen können. Ich danke Ihnen allen, dass Sie mir Ihre Stimme und Ihren Dialekt überlassen haben, und ich danke Ihnen für die vielen interssanten und dialektal geprägten Anekdoten und Einsichten über das Schwäbische und Sächsische. Für die ausführlichen Gespräche und Emails nicht nur über Intonation und Sprache möchte ich mich bedanken bei Gösta Bruce, Peter Gilles, Esther Grabe, Carlos Gussenhoven, Merle Hörne, Bob Ladd, Ineke Mennen, Jörg Peters und Margret Selting. Mit den Tübinger GKlerinnen Katja Jasinskaja und Janina Rado verband mich lange Zeit ein heiteres Baletttanzen, "Es gibt Tee!", und den Potsdamer Kollegen danke ich für ein außerordentlich stimulierendes Arbeitsumfeld. Dank auch an Kristina Vath für die Unterstützung bei der Transkription. Ich möchte mich ganz herzlich bei Sabine Zerbian, Peter Gilles, Hubert Truckenbrodt und Tanja Hüttner bedanken, die Teile dieser Arbeit gelesen und kommentiert haben, and I would like to thank you Cody, and Sam for smoothing my GermEnglish! Nicht zuletzt gebührt meiner Familie großer Dank. Meine Frau Astrid hat mit viel Verzicht, Geduld und reichlich außer-linguistischem Schwung wesentlich zum Entstehen dieser Arbeit beigetragen, tack för maten! Und Liv hat mit ihrem fesselnden "Papa..." H*L H% immer wieder zur gedanklichen Entspannung beigetragen. Danke euch beiden!

„Dies alles sagte er laut und mit ziemlich grober Betonung, in seinem knorrigen Dialekt voller plötzlicher Zusammenziehungen, aber mit einem vertraulichen Blinzeln seiner Augenritzen, welches andeutete: 'Wir verstehen uns schon ...' " Thomas Mann "Buddenbrooks", 1957, Berlin: Fischer, p.286.

Contents

1

Introduction 1.1 Topic of the study 1.2 Aims 1.3 Framework of intonation research 1.4 Structure of the study

1 1 2 3 8

2

The empirical basis 2.1 Overview 2.2 The dialects and recording places 2.2.1 Swabian German 2.2.2 Upper Saxon German 2.2.3 The urban varieties of Stuttgart and Leipzig 2.3 The corpus 2.3.1 Recordings 2.3.2 Informants 2.3.3 Transcription and selection of the data

9 9 9 10 11 12 13 13 14 15

3

Modeling dialect intonation 3.1 Overview 3.2 Accounts of German intonation 3.2.1 Pioneers of intonation research 3.2.2 Autosegmental-metrical studies 3.2.3 Downstep 3.2.4 Linguistic function of intonation 3.2.5 Intonation and Meaning 3.3 Accounts of dialect intonation 3.3.1 Realisational differences 3.3.2 Systemic differences 3.4 An approach of modeling dialect intonation

20 20 20 20 25 34 36 40 41 41 43 45

4

Swabian German Intonation Patterns 4.1 Southern German Intonation 4.2 Nuclear contours 4.2.1 Overview of the data 4.2.2 The rise-fall 4.2.3 The simple rise 4.2.4 The rise-fall-rise 4.2.5 Modification of the rise-fall 4.2.6 Remarks on the simple fall 4.2.7 Boundary specifications

47 47 54 54 57 64 70 73 78 79

VIII

4.3

4.4 4.5

4.2.8 Focus 4.2.9 The phonology of Swabian German nuclear contours Prenuclear accents 4.3.1 Prenuclear falling accents 4.3.2 Prenucleax rising accents 4.3.3 A note on the hat pattern Swabian German dialectal characteristics Summary

80 82 86 86 87 90 90 93

5

Upper Saxon German Intonation Patterns 94 5.1 Upper Saxon German Intonation 94 5.2 Nuclear contours 99 5.2.1 Overview of the data 100 5.2.2 The simple fall 102 5.2.3 The simple rise 109 5.2.4 The fall-rise 116 5.2.5 Modified nuclear contours 118 5.2.6 Boundary specifications 123 5.2.7 Focus 123 5.2.8 The phonology of Upper Saxon German nuclear contours . . . 125 5.3 Prenuclear accents 127 5.3.1 Prenuclear falling accents 128 5.3.2 Prenuclear rising accents 129 5.4 Upper Saxon German dialectal characteristics 132 5.5 Summary 134

6

Realisational differences between Swabian and Upper Saxon German . . . . 6.1 Introduction 6.1.1 The horizontal level of accent realisation 6.1.2 The vertical level of accent realisation 6.2 Alignment of tones in Swabian and Upper Saxon German 6.2.1 Introduction 6.2.2 Analysis procedures 6.2.3 Results of tonal alignment in Swabian German 6.2.4 Results of tonal alignment in Upper Saxon German 6.3 A comparison of accent realisation 6.3.1 Introduction 6.3.2 Analysis procedures 6.3.3 Results 6.3.4 Realisational differences of pitch accents - a model 6.4 Summary

136 136 137 138 139 139 139 141 144 147 147 148 149 149 153

IX 7

A note on intonational meaning 154 7.1 Overview 154 7.2 Intonational meaning 154 7.2.1 A contour based approach 155 7.2.2 A compositional approach 156 7.3 An attempt of analysing intonational meaning 159 7.3.1 The meaning of Swabian German nuclear contours 159 7.3.2 The meaning of Upper Saxon German nuclear contours . . . . 164 7.4 Discussion - Two arguments in favour of a contour approach 168

8

Summary and Conclusion 8.1 Introduction 8.2 Intonational phonology of Swabian and Upper Saxon German 8.3 Comparative intonational phonology: Phonetic realisation . . . . . . . 8.4 Comparative intonational phonology: Intonational meaning 8.5 Conclusion

172 172 173 175 176 177

A Map Tasks

179

Bibliography

185

1

Introduction

„Der Tonfall ist mundartlich ausserordentlich verschieden. Man erkennt jede Mundart sofort an ihrem singenden Charakter. Einen solchen hat jede Mundart, wenn man auch hier monotoner spricht als dort." (Bremer, 1893, 195) 1

1.1

Topic of the study

The above quote addresses the huge amount of variation that one is confronted with in a language, yet the study of dialect intonation has long been a desideratum in German (Auer et al., 2000; Heike, 1983; Wiesinger, 1994, 19). This may be due to the fact that intonologists mainly have focused on theoretical developments of the theory proper, concentrating on national standard languages rather than describe and cover dialectal variation. Traditional dialectologists, on the other hand, have treated the issue of prosody scarcely, since, in this research tradition, it has not been clear which parts of intonation can fulfill the function of distinct linguistic units (cf. Auer et al., 2000; König, 1998, 165). It is a well known fact that intonation not only captures distinct linguistic units but also the gradual phonetic expression of the emotional state of a speaker and pragmatic conversational functions, which makes intonation analysis rather complex (e.g. Fox, 2000, chapter 5 for a general discussion of this aspect). Recently, however, the theory of intonational phonology (cf. Ladd, 1996; Gussenhoven, 2004) has entered the field of dialectology (e.g. among others Barker, 2002; Gilles, 2005; Gilles and Peters, 2004b; Peters, 2006; Truckenbrodt, 2002, 2004), and researchers have become particularly interested in intonational variation, since the intonational diversity of dialects provides a source for several different prosodic phenomena that may advance a development of the theory of intonational phonology (e.g Grice et al., 2000; Grabe et al., 2000; Grabe, 2002; Grabe and Post, 2002; Gilles and Peters, 2004a; Gussenhoven, 2004). Moreover, a growing body of studies of intonational variation allows for comparative and typological research (cf. in particular Grabe, 2002; Gussenhoven, 2004). Following the research tradition of intonational phonology that has been introduced by the groundbreaking study of Pierrehumbert (1980) on American English intonation, the present study is concerned with intonational patterns of two German dialects, Swabian and Upper Saxon German (henceforth SwabG and SaxG, respectively). We address phonological, phonetic and semantic aspects of dialect intonation.

1

"The melody is quite distinct between dialects. One suddenly recognizes each dialect by its singing property. Each dialect is characterized by such a property, however one speaks more monotonous in some places than in others." (Our translation of Bremer, 1893, 195).

2 1.2

Aims

"There is a vast literature on tonal systems in the languages of Africa and Asia, but in spite of many years of dialectological research in Europe, the prosodic systems of varieties of well-known European languages are to all intents and purposes undescribed, while the same is true of most languages spoken elsewhere in the world." (Gussenhoven, 2004, xviii)

In the quote above, Gussenhoven points to the lack of "accurate and theoretically responsible descriptions" (2004, xviii) of European language varieties that would improve our understanding of intonational phonology. The present study attempts to contribute to fill this gap. This study supplements recent studies that have published extensive data on German varieties (in particular Ulbrich, 2004; Gilles, 2005; Peters, 2006; Bergmann, 2006). This study has three aims: 1. To determine the tonal inventory of SwabG (chapter 4) and SaxG (chapter 5) intonation; Our phonological analysis of SwabG and SaxG intonational patterns is based on two assumptions. The first one concerns the peculiarity of dialect intonation, that is, we assume that dialects - at least SwabG and SaxG - are characterized by certain dialect specific intonational features that deviate from Standard German, yet may overlap to some extent. This assumption is shared by the studies of Gilles (2005) and Peters (2006). The second assumption concerns the speech material from which we draw our analysis. We assume that our informants are speakers of the particular dialect (see chapter 2 for a characterization of the informants), and that the intonation patterns gained from our speech corpus reflect dialectal properties. Our results prove this assumption in that we find dialect specific features in both dialects compared to Standard German. That intonational features can be recognized as being dialect specific, has been shown in different perception studies (Gilles et al., 2001; Gilles, 2002; Peters et al., 2002, 2003), though not yet for SwabG and the variety of SaxG that we analyse in our study. 2. To provide phonetic evidence supporting our phonological analysis of SwabG and SaxG intonation, and, based on that, to propose a two-dimensional model that captures the variation of pitch accent realisation between language varieties (chapter 6); Based on the taxonomy of cross-linguistic differentiation of intonation proposed by Ladd (1996), we discuss realisational aspects of intonation, i.e. different phonetic implementation of an identical phonological category. For instance, Grabe (1998a) has shown cross-language differences between Southern British English and Northern Standard German as regards high and low tone alignment as well as truncation or compression of pitch accents. Following the invariance hypothesis of tonal alignment (cf. chapter 6), we show that the different nuclear pitch accents, which we analyse for SwabG and SaxG, differ significantly within a dialect with respect to tonal alignment. We conclude that this phonetic evidence supports our phonological analysis.

3 Further, we attempt to propose a model of pitch accent realisation, which concerns the factors of tonal alignment and pitch excursion size. On the basis of our analysis, we argue that these factors capture the variation of pitch accent implementation between language varieties. 3. To touch upon aspects of intonational meaning (chapter 7); Our discussion of intonational meaning is based on the assumption that nuclear contours convey a certain kind of meaning (Gussenhoven, 1984, 2004). As a matter of fact, our SwabG and SaxG data provide two arguments in favor of this approach, rejecting a compositional account of intonational meaning (Pierrehumbert and Hirschberg, 1990; Hobbs, 1990; Bartels, 1999; Gunlogson, 2001). We argue that in case of complex nuclear pitch contours the analysis of a phrase accent (e.g. Grice et al., 2000) taking scope over an intermediate phrase, for instance, is not appropriate for SaxG. Rather, complex tonal shapes are better understood as being a result of phonological modifications (cf. Gussenhoven, 1984, 2004). The meaning of the resulting nuclear contour is related to the meaning of the corresponding unmodified contour. Interpreting the meaning in terms of a phrase accent would fail. The second argument concerns SwabG where the meaning of any contour appears to be reflected in the postnuclear part of a contour. Given our analysis presented in chapter 4, any SwabG nuclear contour starts with an accentual low tone L*. In terms of a compositional approach, this tone would contribute the same meaning to any contour, which obviously would not make any sense. The nuclear contour approach, on the other hand, captures the whole nuclear contour and, thus, the postaccentual part that contributes to the meaning of the contour. In sum, the analysis of SwabG and SaxG intonation patterns concerns different aspects of intonational variation. On the phonological level, we contrast distinct tonal grammars of the dialects with each other. Apart from a distinct tonal inventory, SwabG and SaxG differ in the use of nuclear contours. On the phonetic level, we show that the two dialects implement certain pitch accents differently. And on the semantic level, we argue that we can apply Gussenhoven's (1984) model of intonational meaning. While the meanings are the same, the two dialects express these meanings with different intonation contours.

1.3

Framework of intonation research

The study of intonation has been and still is carried out in different frameworks where we can identify two major schools. On the one hand, the so called British School has the longest research tradition beginning in the 1920s (e.g. Klinghardt and Klemm, 1920; Klinghardt, 1925) and is often characterized as having pedagogical aims of

4 intonational description (Fox, 2000).2 For a comprehensive overview over the British School system, see Cruttenden (1986); for a brief overview see our review of Hermann Klinghardt's (1925) approach to German intonation in chapter 3, section 3.2.1. The British School system appears to be rather inflexible regarding the differentiation of phonological and phonetic aspects of intonation. Further, for comparative purposes it is less suited. The tone-sequence model of intonation, on the other hand, has emerged from previous American structuralism's approaches to intonation from the mid 1940s (e.g. Pike, 1945; Trager and Smith, 1951), and combines autosegmental and metrical phonology for tonal association with texts (cf. Pierrehumbert, 1980; Ladd, 1996; Gussenhoven, 2004). The tone-sequence model has the flexibility to represent phonological as well as phonetic aspects of intonation. In particular, this framework is suitable for comparative purposes (Grabe, 1998a), since it allows for comparing languages or, as in our case, dialects on different levels where languages/dialects may share certain similarities on one level but may differ on the other. In addition, the framework of the tone-sequence model, or the autosegmental-metrical model of intonation (a term coined by Ladd, 1996), is widely acknowledged in intonation research. For these reasons we base our study on the the tone-sequence model. In the remaining part of this chapter we introduce the basic facts of this model (for an overview see Gussenhoven, 2004, 123ff). The aim of a tone-sequence model is to develop a phonology of intonation analysing it "in terms of melodic correlates of stress and of phrasing" (Pierrehumbert, 1980, 34). The origin of this theory is based on the doctoral dissertation of Pierrehumbert (1980), which aims to provide "an abstract representation for English intonation which makes it possible to characterize what different patterns a given text can have, and how the same pattern is implemented on texts with different stress patterns". It also proposes rules "which map these phonological representations into phonetic representations" (Pierrehumbert, 1980, 2). This model comprises three basic ideas which are listed in the following. 1. A decomposition of intonational tunes into two discrete tonal elements, namely high (H) and low tones (L). 3 The tonal grammar of a language consists, thus, only of these H's and L's, and predicts the language specific set of well formed tonal sequences in a structured way that constitute an intonation phrase (cf. the tonal grammar of 2

3

As a matter of fact, the titles of the studies by Herman Klinghardt explicitly draw on the pedagogical aim as the monographs are intended for teachers and students (e.g. Klinghardt and Klemm, 1920, „Übungen im Englischen Tonfall. Für Lehrer und Studierende"). Later work in this framework has had similar intentions (e.g. von Essen, 1964). Assuming only two levels of tonal events, Η and L, Pierrehumbert improves the former four-level approach originated in the American structuralist tradition (see Pike, 1945; Trager and Smith, 1951). This approach has been criticized due to its rather arbitrary assignment of the four levels of intonation without good reasons for justifying four levels rather than, for instance, three or five (e.g. Bolinger, 1951; Cruttenden, 1986). Further, its fairly opaque set of tonal distinctions has been criticized. For instance, a tonal fall might be indicated by means of a fall from level 4 to level 2, from level 3 to level 2 or from level 3 to level 1. The missing clear-cut distinction of these variants of a fall in relation to intonational meaning has not been fully elaborated.

5 Boundary Tone

P i t c h Accents

Phrase Accent

Boundary Tone

Figure 1.1: Finite-state-grammar of well-formed intonation phrases in American English, from Pierrehumbert (1980, 13).

American English proposed by Pierrehumbert, 1980, in Figure 1.1). The individual tones are related to three different types of tonal units: pitch accents, phrase accents and boundary tones (cf. Figure 1.1). Pitch accents associate with metrically strong syllables of an intonation phrase, and may consist of a single high or low tone, thus being monotonal, or a combination of low and high tones, thus bi- or tritonal. 4 The tone which "lines up with the accented syllable" (Pierrehumbert, 1980, 11) is marked with an asterisk '*'. In accordance with Bolinger (1958), pitch accents represent tonal morphemes. According to Figure 1.1, Pierrehumbert identifies seven tonal morphemes for English. The bitonal pitch accent H*+H - has been rejected in later models (Beckman and Pierrehumbert, 1986; Pierrehumbert and Beckman, 1988).5 Boundary tones associate with the edge of an intonation phrase, and serve as a demarcation. While pitch accents have to be associated with stressed syllables, boundary tones do not need to be; very often, they occur on unstressed syllables since they are assigned for structural rather than prominence reasons, signaling a phrasal edge. To differentiate a boundary tone from other tones, it is assigned with a percentage as a diacritic. A low boundary tone is represented as L% and high one 4

5

Note that Pierrehumbert (1980) and later on (Beckman and Pierrehumbert, 1986; Pierrehumbert and Beckman, 1988; Pierrehumbert and Hirschberg, 1990) assume only monoand bitonal pitch accents; yet Gussenhoven (1984, 2004) assumes tritonal pitch accents for British English, and they are also crucial for the description of SwabG and SaxG (cf. chapters 4 and 5). See Ladd (1996, 273ff) for a discussion of the H*+H~ pitch accent and its rejection.

6 as H% (cf. Figure 1.1). In Standard English and Standard German, boundary tones are monotonal (Pierrehumbert, 1980; Fery, 1993; Grice and Baumann, 2002).6 Finally, the phrase accent has a controversial status in intonational phonology. In the original model (Pierrehumbert, 1980, 5), a phrase accent accounts for the course of pitch between the last pitch accent and the boundary tone. Pierrehumbert and Beckman (1988) changed the analysis of the function of the tone to express a different level of phrasing. A phrase accent is a type of boundary tone and as such it signals the edge of an intermediate phrase, which prosodically represents a smaller intonational unit. Although Ladd (1983) notes the unclear theoretical status of the phrase accent, it recently has been subject to a cross-linguistic review where it is claimed that phrase accents are "primarily boundary tones" but receive a secondary association with "a stressed syllable or some other designated tone-bearing unit" (Grice et al., 2000, 144). As far as German intonation is concerned two contradictory views on the phrase accent exist: whereas some researchers explicitly dismiss the phrase accent - as we do - (Wunderlich, 1988; Uhmann, 1991; Fery, 1993; Grabe, 1998a), others include the phrase accent in the tonal inventory (Grice et al., 1996; Grice and Baumann, 2002; Baumann et al., 2001). For more details about the phrase accent in German intonation see chapter 3. 2. Metrical and rhythmical laws define the prominence relations for all syllables within a phrase (Pierrehumbert, 1980; Ladd, 1996). Using metrical theory provides independent evidence of the prominence relations, i.e. not based on intonational grounds. Tune-text assignment, then, is governed by rules given in (1). (1)

Tune-text assignment rules (Pierrehumbert, 1980, 18) a. If a foot has a pitch accent, any foot of equal or stronger metrical strength in the phrase also has a pitch accent EXCEPT THAT b. There are no pitch accents after the nuclear stress of the phrase.

The metrical grid provides information about stressed (prominent) syllables, and since pitch accents are associated with prominent syllables, the metrical grid predicts the positions of the pitch accents. In (2), the tones associate with the most prominent syllables indicated on level 3 and 4. (2)

Metrical grid of the phrase Peter telefoniert 'Peter is phoning' (Uhmann, 1991, 179); a * indicates primary stress. χ x x x x x x x χ

χ χ χ χ

x x χ χ

level level level level level

4 3 2 1 0

σ* σ σ σ σ σ* (Pe)(ter) (te)(le)(fo)(niert) 'Peter is phoning.'

I

Η* 6

κ

H*L

See Hayes and Lahiri (1991) for bitonal boundary tones in Bengali.

7

Μα ν\ η Η* L- LieFigure 1.2: Six different pitch ranges on the phrase 'Manny' illustrating different degrees of emphasis, from Pierrehumbert (1980, 160).

According to Uhmann (1991), levels 0, 1, and 2 refer to word level stress, 0 representing a syllable, 1 defining potential stress bearing syllables and 2 bearing primary stress, whereas level 3 and 4 contribute to sentence stress. Syllables bearing a prominence on level 2 are the potential docking sites for pitch accents. The most prominent syllable of a phrase indicated on level 4 receives nuclear stress, where the nuclear pitch accent is associated with, i.e. the last syllable of the verb telefoniert 'to phone' in (2). The position of the nuclear pitch accent is mainly determined by rules related to focus and has received much attention (e.g. among others Gussenhoven, 1984; Selkirk, 1984). Preceding syllables bearing prominence on level 3 may be associated to prenuclear pitch accents. In (2), thus, the first syllable of the proper name Peter receives an additional pitch accent. 7 3. A set of rules is responsible for the implementation of abstract phonological tones into an actual phonetic representation, i.e. a Fo contour. Pierrehumbert (1980) assumes two types of rules, evaluation and interpolation. The former "evaluates tones phonetically" and the latter "constructs the F 0 contour between one target value and the next" (Pierrehumbert, 1980, 25). The evaluation of the tones results in concrete, individual Fo values. Interpolation and/or spreading of tones form, then, the corresponding tune of the tonal string. According to Pierrehumbert, the phonetic representation depends on the following criteria: a) The tonal Fo values are computed in relation to the baseline, which is the course of the bottom of the speaker's range over an utterance. Pierrehumbert has found that the final lows of an intonation phrase invariantly reach the bottom of the speaker's range (cf. Figure 1.2). Thus, she assumes the scaling of the Fo values to be dependent on the baseline (Pierrehumbert, 1980, 25f). 8 b) Each tonal target is evaluated from left to right on the basis of its own prominence, and in relation to the phonetic and phonological value of its left tone. 7

8

See Uhmann (1991, 176ff) for the metrical grid and its rules of application for German, and Fery (1988) for rhythmical aspects of phrasing and its tonal implementation. See Ladd (1993) for a more abstract view on the baseline.

8

(a)E Η*

L%

Figure 1.3: Interpolation (a) and spreading (b); shading represents accentuation.

The evaluation process refers to the Fo value of the left tone, but cannot change that value (Pierrehumbert, 1980, 26). c) The variation of peak height depends on the degree of emphasis a speaker is putting on an accent. The pitch peak may vary considerably while the range of lows concentrates on the bottom of the speaker's range (Pierrehumbert, 1980, 20) (cf. Figure 1.2). The second set of rules, i.e. interpolation rules, aim at constructing a concatenated sentence melody, the tune. They operate as soon as a right target receives an Fo value, thus connecting the two targets with each other. Basically, two different sets of rules are assumed. One is responsible for interpolation between two targets, and a second is responsible for tone spreading. The difference between the two is shown schematically in Figure 1.3. In the left panel, interpolation between a high pitch accent and a low boundary tone occurs; the syllables in between axe tonally unspecified. The right panel shows tone spreading, where the trailing low tone of a pitch accent spreads across all of the following syllables until the boundary tone.

1.4

Structure of the study

The second chapter describes briefly our corpus and the method for obtaining it. We provide some background information about the two dialects that we are analysing the intonation of, and give some background information of our informants. The third chapter outlines basic assumptions that we axe following analysing dialect intonation, i.e. the way we are analysing SwabG and SaxG intonation. The fourth and fifth chapters propose an analysis of SwabG and SaxG intonational patterns, respectively. Chapter six presents phonetic evidence that supports our phonological analyses, and proposes a two-dimensional model that captures the variation of pitch accent realisation across language varieties. Chapter seven touches upon intonational meaning where we attempt to show that the Gussenhoven (1984) model of intonational meaning is applicable to SwabG and SaxG. A summary of this study is given in chapter eight where we compare SwabG and SaxG intonation from a phonologically, phonetically and semantically point of view.

2

2.1

The empirical basis

Overview

This chapter provides background information about the dialects which are the subject of analysis and about the speech corpus from which we draw our intonation analysis. Section 2.2 introduces geographical and dialect specific segmental phonological aspects of the two dialects Swabian German (SwabG, section 2.2.1) and Upper Saxon German (SaxG, section 2.2.2). In section 2.2.3, we discuss the choice of the urban varieties of Stuttgart and Leipzig as our recording places. Section 2.3 offers a detailed description of the data elicitation technique (2.3.1), background information about our informants (2.3.2), and an overview over further data handling (2.3.3) such as transcription and the procedure of choosing the speech data for analysis.

2.2

The dialects and recording places

Prom a dialectological point of view, the German speaking area consists of two major dialect areas, Niederdeutsch 'Low German' and Hochdeutsch 'High German' (e.g. Hutterer, 1975; Wiesinger, 1983; König, 1998). Low German is spoken in the Northern part of Germany, while High German comprises Middle and Southern German as well as Austrian and Swiss German. Within Germany, the area of transition between Low and High German is also referred to as Mitteldeutsch 'Middle German', which allows for a division into three major dialect areas (Hutterer, 1975; Wiesinger, 1983). See Figure 2.1 for a classification of the dialect areas in Germany. The two dialects which we analyse in the present study, SwabG and SaxG, belong to the High German speaking area: SwabG is a South-West High German dialect and SaxG an East Middle High German dialect. If we consider the three-way classification of German dialects (see Figure 2.1), we compare a High German with a Middle German dialect. We concentrate our dialect intonation study on High and Middle High German dialects not including Low German dialects for two reasons. First, numerous approaches to Standard German intonation are at hand, and we consider this variety to be closely related to the Low German dialect area. Second, two Low German varieties have been subject of analysis within a large-scale project on German dialect intonation (cf. Auer et al., 2000) where the urban varieties of Hamburg and Berlin are contrasted, among others. For dialect intonational features of Hamburg and Berlin see Auer (2001), Gilles (2001a,b, 2005), and Peters (1999a, 2001, 2002, 2006). Since this study supplements the overview of intonation patterns of German varieties intended by the research project on dialect intonation (Auer et al., 2000), the choice of dialect areas for our study concentrates on urban varieties that have not

10

Figure 2.1: Dialect regions of Germany from Gilles (2005, 66); Stuttgart and Leipzig are added to the original.

been considered within that project. Our analysis, therefore, focuses on Swabian German. Within that project, the Upper Saxon German variety of Dresden has been considered for analysis. Our analysis of Upper Saxon German adds an analysis of a different urban variety, i.e. Leipzig German.

2.2.1

Swabian German

The dialect of Swabian German belongs to the Alemannic group of dialects, which are spoken in the south-western part of the German speaking area including Swiss German (Wiesinger, 1983; Russ, 1989). Dialectologically, SwabG together with Low and High Alemannic are called West Upper German (' Westoberdeutsch'') which belongs to High German (Wiesinger, 1983). Geographically, SwabG is bordered in the north and north-west by the Franconian dialect area, in the east by the Bavarian dialect area, and in the south and south-west by the Alemannic dialect area. According to Ammon and Loewer (1977, 17f) and Russ (1989, 340), the area of SwabG is divided into three subparts, from east to west, into East-, Middle-, and West-Swabian. The urban variety of Stuttgart, which is our object of study, is at the border line between West and Middle Swabian, classified as Middle-Swabian (Ammon and Loewer, 1977; Russ, 1989). A detailed overview over the different isoglosses within the Swabian German dialect area can be found in Haag (1946).

11 Phonologically, SwabG can be characterized by the following segmental features. We only present a few salient dialect features here for illustration. Comprehensive details of Swabian segmental phonology can be found in, for instance, Fischer (1895), Bohnenberger (1928), Prey (1975), Ammon and Loewer (1977), and Russ (1989). - Voiceless, alveolar fricative in syllable initial position, Frey (1975, 14, 25, 29), and no corresponding voiced one as in Standard German (e.g. Fery, 2000); this feature extends to the whole High German area. jxj is pharyngelized in syllable coda position, or before dental-alveolar consort nants, as in ['ε de] 'earth, ground' and ['fi:i] 'for', Frey (1975), Ammon and Loewer (1977, 57), Hiller (1995, 16). ο - /s/-palatalization before stops as in [ fejba ] 'vespers' and ['fejd] 'firm, solid', Frey (1975, 29), Russ (1989, 349); most saliently and frequently in verb inflection, particularly in the second person singular form of the verb sein 'to be' with following / t / deletion (bist): ['bif]. r> - Lack of rounding distinction in high and mid front vowels, as in [uba 1 ] for Std. German über 'above', and ['ble:d] for Std. German blöd 'stupid', Frey (1975, 47), Ammon and Loewer (1977, 39), Russ (1989, 346). As regards SwabG intonation, the only and very brief description to our knowledge is in Frey (1975, 67ff). This is discussed in more detail in chapter 4, section 4.1, reviewing the literature on Southern German intonation.

2.2.2

Upper Saxon German

The dialect of Upper Saxon German is spoken in the eastern part of Middle Germany. Dialectologically, the Upper Saxon German dialect belongs to the East Middle German dialect group ('Ostmitteldeutsch' 1 ) (Wiesinger, 1983). Geographically, the Upper Saxon German dialect area is bordered in the north by the dialect region of the South Mark, in the east by the national border with Poland, in the south by the national border of the Czech Republic, and in the west by the Thuringian dialect (Wiesinger, 1983; Bergmann, 1989). According to Bergmann (1989), the area of Upper Saxon German is a terrace-formation being characterized by many isoglosses that run west-east. The variety of Leipzig, which we are concerned with, is characterized as south-west Οsterländisch (Bergmann, 1989, Osterland meaning Eastland, p.295). The so called terrace-formation of Upper Saxon makes the Upper Saxon German dialect fairly inhomogeneous from a phonological point of view. The following list of segmental features relates to the variety of Leipzig (Albrecht, 1881). For an overview of other Upper Saxon German segmental features see Gelbe (1875), Bergmann (1989), Putschke (1968), and Zimmermann (1992). - Lenition of [ρ, t, k] in initial and medial position, as in [^επίθ] for Std. German Kerne 'pips', Bergmann (1989, 295), Albrecht (1881, 11). - Spirantization of voiced palatal plosive [g] to unvoiced palatal [ς] or, after back vowels, velar fricative [x] in medial position, as in ['za:xa] for Std. German sagen 'to say', or [ze:ga] for Std. German Segen 'blessing', Albrecht (1881, 12).

12

- Spirantization of voiced labial plosive [b] to labiodental fricative [ν], as in [le: var] Leber 'liver', Albrecht (1881, 13). - Lack of rounding distinction in high and mid front vowels, as in ['ri:ban] for Std. German Rüben 'rape', Albrecht (1881, 1, 7f), Bergmann (1989). Regarding SaxG intonation, Gericke (1963) has been concerned with the intonation of Leipzig. Recent studies on Dresden intonation, a neighboring urban variety closely related to Leipzig, have been published (Selting, 2002a,b; Gilles, 2005; Peters, 2006). For a general impressionistic overview see Zimmermann (1998). These studies are discussed in detail in chapter 5, section 5.1.

2.2.3

The urban varieties of Stuttgart and Leipzig

Bremer (1893) points to the influential role of an urban variety with respect to dialectal change and dialect leveling: „Ein anderes Beispiel: Gleich westlich von Braunschweig spricht man statt eines alten % und ü (in Zeit, Haus) Diphthonge. Diese Braunschweiger Landeskinder kommen nun in die Haupstadt, als Soldaten, als Dienstmädchen u. s. w. und hören dort nur von tit und hus. Das macht einen gewissen Eindruck; die heimischen Diphthonge werden unbewusst als ordinär angesehen, und es gewinnt das ϊ und ü nach Westen zu mit der jüngeren Generation an Boden. Hier ist es der Einfluss der Hauptstadt." (Bremer, 1893, XII-XIII, emphasis in the original)1

The prestigious role of an urban variety influences the rural dialects around a larger city. This is what sociolinguistics claims: a larger city can be considered a centre for language change from which innovations spread to other cities and/or the periphery (cf. Kerswill, 1996; Chambers and Trudgill, 1998; Auer et al., 2000). Based on this claim, we assume an urban variety as a centre of a regional dialect. We also assume the intonational patterns of a dialect area to be reflected by such a regional variety. In accordance with Auer et al. (2000), we choose an urban variety as a representative of that particular variety. 2 The urban varieties we chose as the representatives for SwabG and SaxG are Stuttgart and Leipzig, respectively. Another potential urban variety of SaxG would be Dresden, yet the intonation of Dresden has been the subject of analysis already (Auer et al., 2000; Gilles, 2005; Peters, 2006). Stuttgart and Leipzig are comparable with each other with respect to social and cultural aspects. The cities have roughly the same population size and can be considered comparable socio-cultural centres. 1

2

Ά different example: West of Braunschweig, one speaks diphthongs instead of an old i and ύ as in Zeit 'time', Haus 'house'. These children of Braunschweig are coming to the capital as soldiers, maid-servants and so on, and they are only perceiving 'ίζΐ' and ihus\ That gains a certain impression; the native diphthongs are unconsciously judged as rural, thus the i and ύ extends to the west with the help of the young generation. Here, it is the influence of the capital.' Our translation of Bremer (1893, XII-XIII). Of course, there is no homogeneous urban variety. Yet, instructing our informants to speak in their dialectal register, we tried to concentrate on the urban dialect which currently is spoken.

13

Based on the above sociolinguistic claim, the cities presumably have the same status with respect to the regional variety to be analysed.

2.3

The corpus

2.3.1

Recordings

When we are dealing with dialects, some methodological issues concerning the speech data arise. Studies on intonation usually use laboratory speech, in particular read speech (see Lickley, Schepman and Ladd, 2005, for a justification of that procedure). Lickley et al. have concentrated on the change of speaking style leaving a factor like dialect constant. In their analysis of Dutch falling-rising questions, the authors find an identical phonetic implementation of a low phrase tone in read speech and in spontaneously uttered questions recorded in map task sessions (for a detailed review of the map task procedure see below). This result allows the authors to conclude that read speech does not affect the intonation of a speaker as assumed elsewhere in the literature. Thus, the authors argue that analysing intonation contours by means of read speech is a valid procedure. Yet, we doubt this to hold for dialects (cf. also Gilles, 2005, 78ff). Using laboratory speech to study dialects would cause problems since most dialects are hardly codified. This means that we cannot guarantee for a translation from the Standard language into the dialect, since not only the phonology might differ but also morphology, syntax and the lexicon. Still, if we have a suitable codified text that our informants would agree upon as readable in their dialect, we cannot exclude that a change of speaking style (read vs. spoken speech) would not be accompanied with a change in dialect level as well. In particular, our informants have reported that they are able to use different dialectal levels depending on the formal status of the speaking situation. For instance, the informants tend to use a less dialectal level when speaking in formal situations. Some of the informants compared such a formal level of speech with the codified Standard language (Schriftdeutsch). This points to an interference with the Standard that we by no means want to achieve. Given that laboratory speech is not appropriate for our purposes, the question remains which kind of speech material allows for naturalness on the one hand and comparability across speakers on the other. We are of the opinion that map task dialogues fulfill these criteria in that semi-spontaneous speech is elicited. These dialogues are natural, since the participants are free to speak, yet the dialogues are controlled in that participants fulfill the map task referring to the symbols on the maps. The recordings were made at the subjects' homes using a portable Sony DATrecorder and two Sony tie-clip condenser microphones (ECM-TS125). Two subjects were recorded per session. A recording session consisted of two map tasks (Anderson et al., 1991; Claßen, 2000). Each subject acted as the instruction giver and the instruction receiver once, resulting in two map task conversations per session.

14

The recording procedure for the map task is as follows. Two subjects were separated by a shield, thus, participants cannot see each other's map. One of them, the instruction giver, had to describe as accurately as possible a route which was painted on the map. The instruction receiver's task was to draw the route on the map. Both maps contained a starting point and several different symbols, e.g. a caravan, a dragon fly, or a fisherman. However, the two maps differed in three ways: (a) symbols were placed in a different order, (b) not every symbol occurring on one map was given on the other map, (c) symbols carried different names. This procedure caused lively conversations and forced different kinds of questions. The participants were informed that the experiment deals with how exactly information may be coded and transmitted. For that reason, they were told that deviations from the original route will be measured in order to force them to be as precise as possible which means that they have to interact with each other. They were instructed not to gesture, but only to speak with each other. 3 No time limit for the task was given. The map tasks used in the present study are borrowed from Claßen (2000),4 and are given in appendix A.

2.3.2

Informants

We have selected informants who speak the regional variety of Stuttgart and Leipzig, respectively. A prerequisite for speaking an urban vernacular is that informants were born and raised in the respective city. In total, we have recorded 28 speakers, 14 per dialect. However, eight speakers were not considered for analysis, four of each dialect (see below). The age of the speakers ranged from 26 to 66 in case of the SwabG speakers, and from 21 to 74 in case of the SaxG speakers. We recorded both female and male speakers. Table 2.1 displays background information of the speakers. Some of the speakers were not considered for analysis for the following reasons: Speakers Sla and Sib have not referred to the map task fulfilling the task but have calculated the route on the map according to the paper size. This is certainly due to the fact that speaker Sib is an architect. Therefore, we have no data that relate to the symbols of the maps. Speakers S7a and S7b have been excluded since speaker S7b very often has a creaky voice so that we cannot analyse Fo in a reliable way. In addition, the two speakers often overlap their speech which also makes Fo measurements impossible. Although speaker L4b agreed to participate in this investigation, before and after the recording session, she expressed a great amount of uneasiness about being recorded and analysed. In case of speakers L5a and L5b, speaker L5b held a monologue, and speaker L5a hardly interacted. Speaker L5b's monologue sounds like an actor giving a play. This is certainly due to his engagement in an amateur actor's group. 3

4

Studies that were concerned with the methodology of the map task investigated, among others, the influence of eye contact on the performance of the map task (Anderson et al., 1991; Nakano et al., 1997). Nakano et al. (1997) did not find any significant difference between map task performance with and without eye-contact. Our own experience was that subjects were engaged in the task and were thus only looking at their own maps. We observed no eye-contact, and almost no attempt to gesture occurred. I am grateful to Kathrin Claßen who provided me her map task files.

15

Table 2.1: Background information about the informants of the SwabG (Stuttgart) and SaxG corpus (Leipzig); u.f.a. = used for analysis. SwabG - Stuttgart Speaker

Sla Sib S2a S2b S3a S3b S5a S5b S6a S6b S7a S7b S9a S9b

a

Sex

Age

f m f m f f m m f m m f f f

26 31 36 31 31 32 65 56 51 52 51 51 60 66

Profession

student architect housewife mechanician linguist-secretary economist chief of logistics insurance businessman book-keeper technician master house-painter secretary secretary, retired ladies' tailor, retired

u.f.a. -

-

+ + + + + + + + -

-

+ +

SaxG - Leipzig Speaker

Sex

Age

Profession

u.f.a.

m house-painter Lla 26 + m student Lib 27 + f insurance businessman L2a 28 + f interpreter L2b 30 + L3a m 22 student + f 21 student L3b + f economist L4a 58 f 74 L4b competent official, retired L5a f 73 master baker L5b m 65 f kindergarten teacher L6a 57 + L6b f book-keeper, retired 70 + engineer m L7a 58 + m lawyer L7b 58 + a The numbering of the SwabG speakers shows gaps, which are due to plans of having more recordings. -

-

-

-

-

2.3.3

Transcription and selection of the data

2.3.3.1 Transcription The recordings were converted from DAT tape to wav files with a sampling rate of 16 kHz, 16 bit resolution, mono format. The transcription and further analysis was

16 performed in Praat (Boersma and Weenink, 2005). The speech data were orthographically transcribed according to GAT (Selting et al., 1998), which is a system for transcribing conversational speech data. In (1), we give a subset of GAT conventions that we used for transcription of our speech material. The GAT system is mainly based on orthography but additional salient dialectal characteristics may also be transcribed, e.g. (2). In addition, GAT conventions prescribe that words are transcribed in lower case, while upper case refers to accentuation, e.g. (2-b) (cf. (1)).

(1)

A subset of GAT conventions (Selting et al., 1998). 1 [ = (.) (~)> (" ')•> ( ) : ((laughing)) (unintelligible) exAmple eXAMple ((. . .)) —>

(2)

a. b.

each numbered line corresponds to an intonation phrase indication of overlapping speech coarticulation of words or intonation phrases micro pause short, middle, long pause (ca. 30, 60 and 90 msec.) lengthening of sounds indication of paralinguistic events, e.g. laughing an unintelligible part of speech upper case vowel indicates a prenuclear accent upper case syllable indicates nuclear accent omitted part of speech indication of a target sentence discussed in the text

Std. German: A: Hast du einen Wohnwagen 9 GAT, SwabG: A: hasch en WOHNwage

'do you have a caravan?'

Each line in a GAT transcription resembles an intonation phrase. As is the nature of our material, speakers tend not to speak in syntactically well-formed sentences. Intonation phrases may however resemble a syntactically well-formed declarative or question (more about sentence type below). In addition, speakers make slips of the tongue, restart, or produce recipient signals that cannot be counted as an intonation phrase. 2.3.3.2 Selection of speech material From the corpus, we have selected about 200 intonation phrases per dialect for intonation analysis. Since we do not attempt to conduct a corpus study, we did not randomly select phrases or use the whole corpus, but the selection concentrated mainly on intonation phrases which contained enough sonorant material around the accented syllables to illustrate the accentuation pattern properly. We thus excluded phrases that overlapped, phrases that contained disturbances, e.g. interferences that emerge when a speaker is moving, and phrases where hardly any Fo pattern could be detected. We mainly selected intonation phrases that contained the symbols of the map task since these were intended to contain sonorants; (however, many instances of words like Engel 'angel' starting with a vowel showed no reliable FQ pattern at the begin-

17 ning of the first syllable since speakers tend to produce creaky voice in that position which might be a side effect of the glottal stop that obligatorily is realised in these circumstances). The selected intonation phrases were labelled intonationally. Labeling has been based on auditory perception and visual inspection of traces. For technical details of Fo extraction see page 140 in chapter 6. The distribution of the different contours are listed in Table 2.2; the contours are paired between the dialects expressing their functional equivalence. Table 2.2: Overview of frequency of nuclear contours detected in the SwabG and SaxG selected intonation phrases; contours are paired according to their functional equivalence between SwabG and SaxG; + indicates an added tonal morpheme to a base contour, cf. chapter 3. SwabG L*HL% L*HLH% L*H L*HH% L*H+L -

η = 194 76 13 27 35 27 -

SaxG H*L H*LH% L*H L*HH% L+H*L L+H*LH%

η = 201 75 5 21 41 40 19

H*L 15 H*LH% a 1 a Speaker specific, see chapter 4. a

2.3.3.3 A note on sentence types In earlier research, a strong correlation between syntactic structure and intonation has been assumed (cf. for instance Bierwisch, 1966). Empirically, this assumption does not hold, and led to the assumption that sentence mood such as assertion, interrogation, request, or wish correlates with certain intonation patterns (cf. Altmann et al., 1989). Concerning sentence mood, Altmann et al. (1989) draw a clear distinction between interrogative and non-interrogative. This finding brings to mind von Essen (1964) who distinguishes between terminal, progredient and interrogative intonation where the first two appear to belong to the 'non-interrogative' and the latter one to the 'interrogative' mood of Altmann et al. (1989). More recent work on this issue assumes pragmatics to play an important role of interpreting a certain sentence as a declarative or a question (Bartels, 1999; Gunlogson, 2001). In particular, any kind of sentence type can be produced with any kind of intonation. See also Selting (1993), in particular Selting (1995) for a taxonomy of conversational questions, where different levels of linguistics contribute to the interpretation of an intonation phrase as being a question. For yes-no-questions we have shown that they occur with rising and falling intonation in real speech, and the choice of question tune depends on contex-

18

tual effects (Kügler, 2003a, 2004a). Syntax is not the only determiner in establishing whether a sentence is a question or not. For our purposes, we do not assume that a particular sentence type, e.g. a declarative is realised with a certain intonation pattern that only refers to this particular sentence type. There might be a tendency, for instance, for declaratives to be realised with a certain pattern, i.e. some kind of prototypical intonation pattern, but our data shows that there is no one-to-one relation between intonation pattern and sentence type. Yet, we refer to different sentence types, such as declaratives, yes-no-questions, wh-questions or declarative questions. Our labelling of different sentence types was pragmatic, that is, the context in which a certain intonation phrase is uttered determines its status as being a question or a declarative, yet, this classification is not crucial for our analysis but might only guide the reader to some extent. Generally, we classify an intonation phrase as being a declarative if there are no contextual cues that would point to a question. If syntactically complete, this phrase has a verb second position, but note that it need not have a verb at all (ellipsis), e.g. (3-b). The context in (3) determines the declarative interpretation. (3)

An example of an elliptical declarative (3-b) a. A: WAS hast du da 'what do you have there' b. B: en LAMM 'a lamb' (cf. (16-b) in chapter 5)

Accordingly, we distinguish between different kinds of questions, the most obvious ones are yes-no-questions and wh-questions. The former are syntactically marked with verb inversion, e.g. (4-a), and the latter contain a wh-word, e.g. (3-a) above. A declarative question refers to a yes-no-question that is not syntactically marked by verb inversion, i.e. a declarative question does not contain any interrogative marker (e.g. Haan, 2001; Gunlogson, 2001), as in (4-b). (4)

Two kinds of yes-no-questions a. A: hasch du den da DRAUFzeichnet 'do you have it drawn on your map?' b. A: drEhende Windmühle hast du NICHT 'don't you have rotating windmill' B: nein 'no'

(cf. (14-a) in chapter 4)

(cf. Fig. 5.18d in chapter 5)

A final note on declaratives must be made. In case of declarative structures that do not signal finality, we distinguish two types that signal continuity: 5 (i) a list continuation, and (ii) a sentence or turn continuation. Type (i) of continuation does not refer to what intonologists have been called list intonation. The term list is borrowed from interactional linguistics (cf. Brinker, 2000). In this sense, a list refers to a conversational activity, where list items are sequentially 5

Our distinction here is equivalent to the distinction made by Gilles (2005, 52ff, 112): cont-I corresponds to Gilles' "parallel-ranked", and cont-II to "successively-ranked" continuation. Gilles provides syntactic, semantic and pragmatic criteria that determine continuity.

19

embedded and may be interrupted by interlocutors. Further, list items are generally produced syntactically parallel, and their intonation functions as a turn-holding device (Selting, 2003, 2004). Lists in terms of traditional intonation research, on the other hand, refer to single intonation phrases where the phenomenon of downstep, reset and final lowering is analysed (cf. Beckman and Pierrehumbert, 1986; Fery, 1993, for English and German respectively). Selting (2003) distinguishes between open and closed lists, and assigns lists analysed in intonational phonology as an example of closed lists. For examples of continuation-I, see (12) in chapter 4 (page 68), and (8) in chapter 5 (page 110). Type (ii) of continuation concerns instances where speakers intend to complete complex sentences. This concerns subordinated clauses that start with conjunctions like 'if', 'although', 'since', and so on. In these cases, it is a sentence completion. This type of continuation also concerns turn completion, when speakers phrase a sentence into a sequence of succeeding intonation phrases. Otto von Essen has called an intonation pattern that coincides with that part of an utterance to be completed as progredient ("weiterweisende (progrediente) Aussprüche", von Essen, 1964, 37). For examples of continuation-II see (13) in chapter 4 (page 68) and (10) in chapter 5 (page 113).

3

3.1

Modeling dialect intonation

Overview

In chapter 1, we briefly introduced the framework of intonational phonology in which the present investigation is carried out. This chapter outlines different approaches to German intonation within that framework, and intends to propose an autosegmentalmetrical system which allows for the analysis and comparison of Swabian and Upper Saxon German intonation patterns. We basically follow Gussenhoven (1984, 2004), with additional assumptions on German intonation based on Fery (1993) and Grabe (1998a), and assumptions for dialect comparison based on the taxonomy of different levels of intonational differences between languages proposed by Ladd (1996, 199ff). The review of German intonation in section 3.2 starts with early work on German intonation in section 3.2.1. In section 3.2.2, we discuss two distinct autosegmentalmetrical accounts of German intonation, outline the differences in tonal structure and tonal representation between the two, and discuss how these aspects of intonation relate to our system. Section 3.2.2 further discusses aspects of linguistic function and meaning of intonation that are relevant to our system. Section 3.3 is concerned with different levels of intonation that may differ between languages and/or dialects. In particular, we discuss aspects of realisational and systemic differences of intonation. The system that we propose for the analysis and comparison of dialect intonation is summarized in section 3.4.

3.2

Accounts of German intonation

3.2.1

Pioneers of intonation research

It is noteworthy that some of the basic understandings of intonation have already been introduced by researchers of the 19th and the early 20th century while some issues developed further in recent intonation research become much clearer on the basis of earlier misunderstandings. Otto Bremer, for instance, has pointed to the distinction of stress ("Betonung") and accent ("Akzent") (Bremer, 1893, § 181 ff.). For Bremer stress is observed by the strength of the voice ("Stärke des Stimmtons" §§ 180 -181) whereas accent corresponds to the height of the voice ("Tonfall (Höhe der Stimme)" § 194). Similar to our understanding of the distinction between stress and accent, Bremer assumes stress to be a lexical property, i.e. words do have word stress, while the accent may change, i.e. the shape of the accent might be high or low depending on the meaning of the phrase (§ 196).

21

A further insight of the relation between the speech melody of a sentence and the sentence itself is made by Bremer comparing different sentences with the same melody: "Im einfachen Aussagesatz ist unser Tonfall normalerweise durchaus an die Betonung gebunden." (§ 199).1 Bremer emphasizes the fact that the stressed syllable is bearing the accent, and the accent, thus, comprises both the phonetic correlate of stress, i.e. the strength of the voice, and the phonetic correlate of accent, i.e. the height of the voice. However, Bremer's observations refer to the domain of words not to sentences since he assumes the sentence melody to be an issue of syntax (§ 196). Based on our current knowledge of intonation, this assumption appears to be a mis-interpretation of the role of intonation in relation to phonetics and phonology. In any case, Bremer (1893, 195) describes the canonical intonation pattern of a German declarative sentence exhibiting a high accented nuclear syllable and a falling accent. 2 Hermann Klinghardt introduced the first model to cover the linguistics of intonation (Klinghardt, 1925, 1927; Klinghardt and Klemm, 1920; Klinghardt and Olbrich, 1925).3 Klinghardt particularly emphasizes the need of intonation for second language learning since the intonation appears to differ between languages. Accordingly, it is not sufficient to learn only the sounds of a foreign language but also the intonation (Klinghardt and Klemm, 1920, 3f.). The work by Barker (1925) and von Essen (1964) continues in the tradition introduced by Klinghardt in providing a pedagogical oriented comprehensive overview of German intonation patterns. Klinghardt's model explicitly distinguishes emotional, paralinguistic, and linguistic parts of intonation. The domain of speech melody is the stress-group ("Sprechtakt"), which he defines as "kleine Wortgruppen, die man verständigerweise nicht trennen kann" (Klinghardt, 1925, 7).4 Stress groups may be suspensive ('Veiterweisend") or final ("abschließend") (p. 9). The former is characterized by a rising tone movement ("steigende Tonbewegung"), and the latter by a falling movement ("fallenden Ausgange") (ibid.). The internal structure and the length of a stress-group may vary. On average, Klinghardt counts eight to twelve syllables per stress-group in prosaic texts. With respect to the internal structure, Klinghardt proposes three parts that a stress group consists of: a pre-head "Auftakt", a head "Taktkopf', and a tail "Abtakt" (p. 20). 5 The head of a stress-group is obligatory, the other two are optional elements. An example of a final stress-group is given in (1). The first part in parenthesis is given for reasons of understanding the sentence and represents a suspensive stress-group. 1 2

3

4 5

'In a simple declarative sentence, our accent normally is determined by stress.' "Hier gilt als Regel, dass die starkbetonte Silbe den Hochton trägt, die nebenbetonte den Ebenton, die schwachbetonte den Tiefton. [...] der Worttonfall ist sinkend" (Bremer, 1893, 195). 'The rule here is that the strongly stressed syllable is bearing the high tone, the secondary stressed syllable the level tone, the weakly stressed syllable the low tone. [...] the accent is falling.' The theory of intonational description developed by Klinghardt (1925) is the source of the so called British School of intonation (cf. also Fox, 2000, 278). 'small word groups that one cannot separate reasonably'. The English translation of Auftakt, Taktkopf, and Abtakt is given in British School terms (cf. Fox, 2000, 278f). Note that the British School refers to the last accent as to the nucleus, which is included in Klinghardt's Taktkopf.

22

Strongly stressed syllables, as Klinghardt defines the accented syllables, are indicated by a stress mark "' preceding the syllable. Unstressed syllables prior to the head of the stress-group belong to the pre-head and are delimited by a square bracket']'. The head comprises all syllables from the first until the last accented ones. All syllables after the last accented syllable belong to the tail. (1)

Internal structure of a stress-group (Klinghardt, 1925, 26) (die Belagerung Jerusalems) (stress-group 1 ) steht im] 'Vordergrund dieser 'spannenden Er zählung. Auftakt Taktkopf Abtakt pre-head head tail 'The siege of Jerusalem is in the foreground of this exciting narration.'

Intonation curves are graphically represented by a plot of circles differing in height from an average pitch level indicated by a horizontal line. The final stress group in (1) is shown in Figure 3.1 to illustrate Klinghardt's system. Every dot represents one syllable; smaller dots represent unstressed, bigger dots, stressed syllables. Klinghardt emphasizes that the distance between dots does not resemble any relationship between the syllables but is due to graphical and practical reasons (p.21). •

w ·

· t_

9

·

steht im Vordergrund dieser spannenden Erzählung Figure 3.1: Graphical representation of (1) according to Klinghardt (1925, 26).

In German stress groups, Klinghardt observes a systematic downtrend. Syllables in the pre-head show the same downtrend as in the head, while syllables occurring in the tail level out (cf. Fig. 3.1). As an exception from this general pattern, Klinghardt observes accented syllables that block the downtrend, i.e. an accented syllable that rises to a higher pitch than the downtrend line. According to Klinghardt, these accents occur if a stress-group contains more than one important word. It appears to be a case of broad focus that strikingly occurs very seldom in Klinghardt's data. As a student of Klinghardt, Barker (1925) applies the model of her teacher, and provides a detailed and far more comprehensive study of German intonation than Klinghardt (1925), since Klinghardt has concentrated on theoretical issues concerning the model and its application for comparative purposes contrasting English, French, and German. Thus, Klinghardt's sections about the individual languages exhibit a rather summarizing character of each of the languages' major intonational properties. Marie Barker provided the first detailed "Handbook of German Intonation" with several different prose texts transcribed intonationally. Following Klinghardt, Barker differentiates between two basic intonation phrase types - 'tone-group' in her terminology - a suspensive and a final one. The former "indicates that the sense is not

23 completed or that more is to follow", and the latter "indicates completion" (Barker, 1925, 1). Each tone-group consists of a "final strongly stressed syllable" (ibid.), i.e. a nuclear accent in current terminology. The nuclear accent in a suspensive tonegroup is characterized by a rise in pitch, and in a final tone-group by a fall. Given the tone-groups' different functions and their association with different tones, Barker inherently expresses a correlation between a tone and its function. Within each tonegroup, further words may be accented, thus acknowledging prenuclear accents. These may differ in form. Strongly stressed syllables may either descend in the course of pitch or they may receive higher pitch. Consider an example from (Barker, 1925, 5). In (2), a complex sentence is divided into tone-groups indicated by a super-scribed number. See Figure 3.2 for a graphical representation of (2). The first four tone-groups are suspensive ones, that is, the sentence is not completed, which is indicated by a final rising accent, respectively. The difference between the two alternative pronunciations in Figure 3.2 consists in prenuclear accent structure. According to Barker (1925), the prenuclear peaks in the upper panel stand out from the general downtrend, which makes the accent a rising one, while in the lower panel, the prenuclear accents are in the course of the general downtrend which makes them falling ones. (2)

Während der Bewegung des Zuges1 | ist das Offnen der Wagentür, 2 | das Betreten der Plattformen, 3 | und das Ein- oder Aussteigen4 | verboten. 5 || (Barker, 1925, 5) 'During movement of the train, opening of the door, standing outside, and getting in and off is forbidden.'

* · 1

•· 1

Ȋ

w

4

·

I

Sjj

· •»«

Figure 3.2: Graphical representation of a complex sentence divided up into five and, alternatively, four tone-groups, from Barker (1925, 5).

The "Handbook of German Intonation" (Barker, 1925) is concerned with prose texts. Thus, the tone-group differentiation refers to written speech only. The general pattern of openness which is expressed in suspensive tone-groups can, however, also be found in conversation where openness functions also as a turn-holding device (see our data analysis in chapters 4 and 5, and Selting, 2003; Gilles, 2005). Barker (1925) essentially proposes two distinct accents, a falling and a rising one. In nuclear position, a falling tone expresses finality, a rising one openness. In prenuclear position, either a falling or a rising accent may occur. Barker's main contribution to German intonation is her acknowledgment that "the final strongly stressed syllable of a final tone-group often has the highest pitch" (Barker, 1925, 2 and 20, emphasis in the original), which is not shown in Figure 3.2. This means that Barker has noticed

24 the variation that concerns falling nuclear accents in German (see next section). Yet, Barker ascribes the high accent to conversational speech. Barker does not mention, however, several other accent patterns, or variants of accent patterns, that are described to occur in German in later work (e.g. the early peak or a rise-fall; Fery, 1993; Grice and Baumann, 2002). The early work of Klinghardt (1925) and Barker (1925) reflects the main patterns of German intonation, though only at a basic level. A more detailed description of German intonation has been provided by Otto von Essen (1964) following the tradition of intonational work laid out by Klinghardt. Otto von Essen differentiates between three basic intonation phrase types (cf. Figure 3.3). The suspensive and the final types are known from Barker (1925) and Klinghardt (1925). The interrogative type is introduced by von Essen. 6 He discusses two theoretical issues, the first concerns positional effects of the accent within a sentence, and the second one, details the presentation of different sentence types and their tonal construction.

a) terminal:

**]""*·

' t

b) progredient:

·

β

J

* · « . £ · ·

c) interrogativ:

·

^ J

" " ' f * ^

Figure 3.3: Summary of tonal structures proposed by von Essen (1964, 64).

With respect to the position of the accent in the phrase, von Essen compares accents on sentence final syllables with non-final ones. Given an accented syllable with high tone, pitch falls on the post-accentual syllable to the bottom of the speaker's range ("Lösungstiefe" von Essen, 1964, 21) and remains flat until the end of the phrase. If the accented syllable is phrase-final, the fall is realised on that syllable. With respect to sentence type, von Essen relates the three basic tonal patterns to distinct sentence types. Terminal intonation is used in statements, requests, exclamations, addresses, wh-questions, indirect speech, and in the second part of alternative questions; progredient intonation is used in incomplete parts of speech, in particular parts of sentences, subordinate clauses, and the first part of alternative questions; and interrogative intonation is used in yes-no-questions, echo-questions, rhetorical questions, and as variants of statements, requests and yes-no-questions if they are used expressing a warning or threat (von Essen, 1964, 65f).

6

Since Barker (1925) concentrated on prose texts, in her study of German intonation questions have not been an issue. In the handbook, only five question occur within a conversational passage (Barker, 1925, 67ff). The questions are treated as beiiig uttered in suspensive tone-groups.

25

3.2.2

Autosegmental-metrical studies

In the framework of the autosegmental-metrical theory of intonation, German intonation has been modeled in two distinct accounts. The main proponents of one line of research are Fery (1993) and Grabe (1998a) with the precursors Wunderlich (1988) and Uhmann (1991). The other line of research has emerged in the work of the German ToBI system (e.g. Grice et al., 1996; Grice and Baumann, 2002).7 In the following sections, we outline a brief summary of the accounts of German intonation, and attempt to identify the crucial differences between the two groups of intonation research that concern tonal as well as representational aspects of intonation. See also the discussion in Grice and Baumann (2002, 2) about tonal differences between the two distinct accounts.

3.2.2.1 Overview of German intonation This section briefly summarizes the accounts of German intonation that have been carried out within the autosegmental-metrical framework of intonation. The studies reviewed here are Wunderlich (1988), Uhmann (1991), Fery (1993), Grabe (1998a), and GToBI based on Grice and Baumann (2002). The account of Wunderlich (1988) proposes six different accents which are presented in Figure 3.4. Some of the patterns appear to be closely related, and it is thus questionable whether the accent inventory should contain six accents. 8 In particular, accent D and Ε (cf. Fig. 3.4) might be tonally equivalent if analysed early in an intonation phrase, thus, avoiding the influence of tonal crowding at intonation phrase boundaries. The same might be true for accent C and Β (cf. Fig. 3.4). Uhmann (1991) proposes four distinct pitch accents for German, two bitonal (H*+L, L*+H), and two monotonal (H*, L*). The monotonal ones, however, are restricted to prenuclear position. This observation is reflected in Wunderlich's accent F, and is also in line with Fery (1993) and Grabe (1998a). According to Uhmann, the distinction between mono- and bitonal accents is justified, since monotonal accents do not affect post-accentual syllables while bitonal accents do. In addition, Uhmann proposes a low (L%) and a high (H%) boundary tone for German, following the approach of Pierrehumbert (1980) for English. The most comprehensive work on German intonation is by Fery (1993) who proposes four distinct nuclear contours, and two further nuclear tones, the stylized contour (H*M) and the early peak accent (H H*L). A simple fall (H*L) and a simple rise (L*H) constitute the core accents in German. This assumption corresponds to the earlier descriptions of Klinghardt (1925), Barker (1925), and von Essen (1964). 7

8

ToBI stands for 'Tone and Break Indices', and provides conventions for transcribing intonation. It was originally developed for American English (Silverman et al., 1992; Beckman and Ayers-Elam, 1997) on the basis of Pierrehumbert (1980). A ToBI system has been developed for different languages (for a recent overview see Jun, 2005), among others for German (Grice et al., 1996; Baumann et al., 2001; Grice and Baumann, 2002; Grice et al., 2005). See Grabe (1998a, 24-25) for a detailed review and critique of Wunderlich (1988).

26 Accent Pattern

schematical

and

tonal representation

Α Gipfelakzent 'Peak accent'

H*

Β Brückenakzent 'Bridge accent'

Η* Η L*

C Fallend-Tiefakzent 'Falling low accent'

%H L*

D Tiefakzent-Steigend 'Low accent-rising'

L* H%

Ε Echoakzent 'Echo accent'

L* Η (H%)

F Linker Brückenpfeiler 'Left bridge pier'

Η* Η

Figure 3.4: Accent patterns of German according to Wunderlich (1988, 11).

Fery further assumes that the trailing tone of an accent spreads until the end of the intonation phrase, and accounts thus for the postnuclear pitch patterns. Given this analysis, Fery rejects the need for boundary tones, except for the fall-rise contour, which is a combination of the simple fall and a facultative high boundary tone (H*L H%). The fourth nuclear contour, the rise-fall, led Fery to assume a tritonal movement, i.e. a low accent followed by a bitonal trailing tone (L*HL).9 The simple fall and simple rise may also occur in prenuclear position. Based on the theory of tonal linking (Gussenhoven, 1984), Fery assumes three surface variants of underlying bitonal accents, thus assuming two layers of tonal representation, an underlying and a surface layer. An underlying prenuclear accent may either be realised as a completely linked tone, where the trailing tone is deleted; or it may be realised as a partially linked tone, where the trailing tone is shifted rightwards prior to the next pitch accent. If no tonal linking is at issue, two fully realised bitonal pitch accents occur in sequence on the surface. Grabe (1998a) examines German accents phonetically comparing the tonal phonology of English and German. The analysis considers nuclear as well as prenuclear falling and rising accents following mainly the approach of Fery (1993). Grabe also assumes two layers of tonal representation, where surface tonal variation is accounted for by tonal linking rules. Yet, Grabe also extends their domain of application to nuclear accents. In addition, Grabe considers downstep as an accent accommodation effect, i.e. downstep is considered to be categorical. Grabe presents evidence that falling accents show a gradual behavior with respect to the amount of downstep. In German, partially and completely downstepped accents occur; a completely downstepped accent lacks high pitch on the accented syllable. At first glance, a completely downstepped accent might therefore be misleadingly analysed as an early fall (Grabe, 1998a, 90). See section 3.2.3 for a discussion about downstep. 9

See Grabe (1998a), Gussenhoven (1994), and Ladd (1994) for a critique of this approach.

27 Grabe's analysis of falling accents reveals that the surface realisation depends on the position of the accent in the intonation phrase (cf. also Goldbeck and Sendlmeier, 1988). Non-final falls align the high accentual tone at the right edge of the syllable rime, whereas phrase final falls align at the left edge of the syllable rime (cf. Figure 3.5). Any pitch shape prior to the fall is considered to be phonetic in nature, i.e. falls may be preceded by an onglide or a pitch plateau from where the fall starts. For rising accents, Grabe observes more variation. A common characteristic of a rise is that the postaccentual syllable is higher than the accented one, i.e. the accentual peak is realised in the postaccentual syllable. The realisation of the accentual low varies within the accented syllable.

(a) Non-final position

(b) Final position

(c) Final position

with voiced onset

with voiceless onset

Η

/ g el gelben

b

ψ/

/ν d Wolf

17

Η

Λ £U

ς/

Teich

Figure 3.5: Accent shapes of nuclear falling accents (H*+L) in non-final and final position, from Grabe (1998a, 75).

German intonation has been modeled within the ToBI framework, i.e. German ToBI (GToBI, e.g. Baumann et al., 2001; Grice and Baumann, 2002). In GToBI, six distinct pitch accents are assumed, two of which are monotonal and four bitonal. As with ToBI (Beckman and Ayers-Elam, 1997), GToBI is a surface oriented approach to label intonation, therefore, no underlying accents are assumed. Contrary to the above approaches, GToBI allows for leading and trailing tones. As a result, the accent inventory contains more accents than, for instance, Fery (1993). In addition, GToBI proposes two levels of phrasing, the intermediate phrase and the intonation phrase. An intermediate phrase boundary is signaled by a phrase accent, whereas an intonation phrase boundary is signaled by a boundary tone. Any intonation phrase consists of at least one intermediate phrase. Thus, if the phrase accent and the boundary tone have identical tonal values, their boundary specifications collapse into a single label, e.g. a low phrase accent and a low intonation phrase boundary are labelled as L-%. To sum up this brief overview of German intonation, the nuclear contours of the approaches reviewed are summarized in Table 3.1 on the next page. The table is adopted from Grice and Baumann (2002, 24). 10

10

See also a recent proposal of Standard German intonation (Peters, 2006), which we do not include here, since Peters' proposal is mainly based on previous work on German intonation which partly has been discussed above.

28

Table 3.1: German nuclear contours, adopted from Grice and Baumann (2002, 24).

Falling

la lb

Rising-Falling (delayed peak)

2

Rising

3a 3b 3c

Plateau

4

Falling-rising

5

Early Peak

6a 6b

Stylized downstep

Wunderlich Η* L

L* Η H% L* H%

Uhmann H * + L L%

Fery H*L

GToBI H* L-% L + H * L-%

L * + H L%

L*HL

L * + H L-%

L * + H H%

L*H

L * + ( H ) H-~H% L* L-H% ( L + ) H * H-~H%

L*H

( L + ) H * H-(%)

H*L H%

( L + ) H * L-H%

H+H*L

H+!H* L-% H + L * L-%

H*M

( L + ) H * !H-%

H * + L H% %H L* L

7

3.2.2.2 Two distinct approaches of German intonation This section attempts to identify the relevant differences between the two distinct approaches of German intonation. The differences concern tonal and representational aspects of intonation. The tonal aspects concern the internal structure of pitch accents (leading vs. trailing tones), the relevance of phrase accents, and the types of intonation phrase boundary tones. The representational aspects concern the question whether an underlying level of tonal representation is assumed or not. The discussion starts with the individual tonal aspects that differ. Tonal aspects — the internal structure of pitch accents While Fery (1993) and Grabe (1998a) assume only left-headed pitch accents, i.e. an accentual tone followed by a trailing tone, the GToBI system assumes both left- and right-headed pitch accents.11 Any right-headed pitch accent contains a leading tone before the accentual tone. See (3) for a comparison of leading and trailing tones, and Grice (1995, 120ff) for a detailed discussion about theoretical considerations of the internal structure of pitch accents. (3)

A comparison of leading (3-a) vs. trailing (3-b) tones (T = tone). a. b.

T+T* T*+T

Based on the assumption of leading tones, GToBI proposes two distinct nuclear falling accents (H*L— and L+H*L—), whereas Fery (1993) and Grabe (1998a) assume only 11

In case of the early peak (H H*L), Fery (1993) acknowledges that the pitch shape prior to accented syllable plays a crucial role for that particular accent category.

29

one (H*L). A comparison of the two contours is given in (4), where (4-a) occurs in a neutral statement, broad focus, and (4-b), in a contrastive statement, narrow focus (Grice and Baumann, 2002, 19). For the notation of focus see section 3.2.4. According to Fery (1993), there is no tonal distinction between narrow and broad focus (cf. (4-a) and (4-b)). In GToBI, however, it is assumed that the nuclear high accent preceded by a leading low tone signals a kind of contrast, thus narrow focus (cf. (4-b)), in contrast to a neutral way of uttering a declarative sentence (cf. (4-a)). The role of the phrase accent in the GToBI labelling of (4) will be discussed later (see below). (4)

Comparison of nuclear falling contours between GToBI and Fery (1993) according to Grice and Baumann (2002, 19, Table 1). Example (4-a) is from Fery (1993), (4-b) is from von Essen (1964). Tonal association is our adaptation. a. GToBI H* L-%

I

/I

Mein ZAHN tut WEH Fery b.

GToBI

Fery

'My tooth is hurting.'

H*L L+H*

L-%

Schon der VerSUCH ist STRAFbar H*L 'Already the attempt is criminal.'

Fery (1993) and Grabe (1998a) assume that the pitch shape prior to the accentual high is phonetic in nature. That means that the amount of pitch change of a potential onglide to the accentual high does not affect perception of this accent as a falling accent category. Grabe has conducted a discrimination test on the basis of items classified as a falling accent from her corpus. The result was that one cannot distinguish perceptually between falls with or without an onglide. This test seems to support Fery's analysis; however, if the GToBI assumption that different accents signal narrow and broad focus holds, then Grabe's results of her small perception experiment does not necessarily contradict the GToBI assumption, since Grabe did not control for focus structure in her experiment. A recent proposal on Standard German assumes a leading low tone in case of narrow focus as a modification of a high or falling nuclear accent (Peters, 2006). On the other hand, the GToBI assumption of different tonal signaling of broad and narrow focus is proposed without further exploration. While the two tones are assumed to differ functionally, no attempt has been made to differentiate this pattern formally. The example given in Grice and Baumann (2002, 21) does not show the phonetic implementation of a leading tone, and no description of alignment or association is given. According to the theory of intonational phonology, a leading tone must be associated with a syllable preceding the accented one (cf. our tonal adaptation in (4)). The phonetics of Grice and Baumann's example, however, rather suggests an

30 association of the leading low with the accented syllable. Further, this functional distinction has been claimed only for falling accents (cf. contours la and lb in Table 3.1 above). If this is the case, it is questionable why GToBI assumes the leading low in other contours as well where a high accent occurs (cf. the contours with a leading low tone in parenthesis in Table 3.1 above). Why should the onglide in case of a falling accent be a phonological relevant tone, and in all the other cases just phonetic variation, i.e. a possible way to express the general contour in two ways? From this discussion, we conclude that it is obviously an open debate whether or not Standard German may be modeled in terms of leading tones. On the one hand, the phonetic nature of the onglide has not comprehensively been proved perceptually. On the other hand, the functional differentiation of the two falling accent shapes has not been explored in detail. Since we do not attempt to model Standard German we leave this discussion here. Assuming a low leading tone, GToBI thereby also assumes a high leading tone ( H + L * ) . However, Grabe (1998a) argues that a H + L * accent can be accounted for in terms of a completely downstepped accent (!H*+L). Grabe has experimentally shown that the amount of downstep in German is gradual, ranging in a continuum from partially to completely downstepped. Grabe analyses downstep as a categorical phonological adjustment. Tonal aspects — the case of the phrase accent A second crucial distinction between the two approaches of German intonation concerns the presence or absence of phrase accents (cf. chapter 1, section 1.3 for the status of the phrase accent in intonational phonology). Although Fery (1993) acknowledges different levels of phrasing, an intermediate phrase boundary is proposed to be left tonally unspecified. Instead, spreading of the trailing tone until the end of the phrase accounts for the course of pitch between the accent and the boundary (cf. (4) above). In general, Grabe (1998a) follows Fery's analysis. In particular, Grabe rejects the phrase accent for two reasons: first, due to "the lack of evidence for the intermediate phrase in German" (Grabe, 1998a, 47); second, since Grabe's aim is to compare the intonation of German and English and since she has no evidence for the phrase accent in English, her analysis also postulates tonal spreading as in (4). Contrary to Fery and Grabe, GToBI proposes the phrase accent for German intonation. This proposal was based on an experimental investigation of the alignment behavior of the low trailing tone in falling accents where the number of unstressed and word stressed syllables after the nuclear accent have been systematically varied (Benzmüller and Grice, 1998). The main result was that the fall in pitch and, thus, the alignment position of the low trailing tone depends on the position of the following word stressed syllable. Since the fall shows a pattern independent of the accentual high but dependent on the structure of following syllables, Benzmüller and Grice conclude that the low tone does not belong to the pitch accent but receives its own independent status as a phrase tone. Additional evidence for the phrase tone comes from an investigation of the Eastern European Question Tune (Grice et al., 2000). Given that phrase accents may have secondary association (Pierrehumbert and Beckman, 1988), Grice et al. argue for the existence of a phrase accent in intonational phonology. In particular, the

31 authors assume phrase accents to be edge tones that may secondarily associate at certain structural points in the phrase. In a comparison of Hungarian, Romanian, Greek and German, the authors postulate t h a t the phrase tone associates secondarily at postnuclear stressed syllables in Standard Romanian, Standard Greek, and Standard German, whereas it associates secondarily to syllables near the phrase boundary independent of the prominence of that syllable in Hungarian and Cypriote Greek. W h a t remains to be clarified, however, is the nature of the phrase accent's primary association, as Grice et al. (2000, 181) have noticed. Further, the relevance of the phrase tone in German still remains open to our view. Although Benzmüller and Grice (1998) experimentally measured the alignment of the postaccentual low tone, the authors recorded only two speakers. From a methodological point of view, this procedure does not hold any statistical significance. In particular, since the authors observed a certain amount of variation concerning the actual alignment of the falling pattern.

Tonal aspects — boundary specifications As far as boundary tones are concerned, Fery (1993) and Grabe (1998a) assume that the tonal inventory of Standard German contains only a facultative high boundary tone. Fery observes no specific pitch movement at phrase boundaries in many cases and argues that the trailing tone of nuclear bitonal accents is responsible for I P boundary specifications assuming that the trailing tone spreads until the end of an intonation phrase (5-a). Thus, in an overall falling contour containing a nuclear falling pitch accent (H*L) as in (4) above, a low pitch until the end of an intonation phrase is a reflex of the trailing L tone. 1 2 T h e facultative high boundary tone is assumed to occur in a tritonal nuclear contour (5-b) where it accompanies a falling pitch accent as in (6). Generally, this accent pattern goes with questions, as in (6), but Fery also mentions that this accent pattern is "used to express a threat" (Fery, 1993, 91). (5)

a.

Spreading of a trailing tone σ σ σ ... σ IIP

I M^

T*+T b.

Association of a high intonation phrase boundary tone

I

σ

σ σ ... σ IIP

T*+T (6)

Η% Η* L

Η%

I ΚI

Wo hast du das A U T O geparkt ? 'Where did you pull in the car?'

(Fery 1993, 92)

Grabe (1998a, 47ff), by comparison advocates the position that phrase boundaries do not need to be marked tonally, since she assumes that boundary tones are "language12

Fery (1993, 74) also notes that the end of falling contours is affected by final lowering.

32

and dialect-specific" (Grabe, 1998a, 48). The assumption of an obligatory low and high boundary tone as put forth in Pierrehumbert (1980) would yield wrong predictions comparing for instance, the English variety of Northern Irish English with American English (cf. Grabe, 1998a, 48). In Irish English, a low boundary tone has its correlate in a clear downward pitch movement (e.g. Jarman and Cruttenden, 1976; Cruttenden, 1995; Dalton and Ni Chasaide, 2003, 2005), which does not exist in American English. To deal with this comparative asymmetry, Grabe concludes that the approach of Pierrehumbert (1980) should dispense with a low boundary tone. To indicate an unmarked boundary, Grabe labels it with 0%. Note that 0% expresses a phonetic level of tonal transcription, i.e. it only signals that no phonological tone is assumed to occur at the phrase boundary. Fery's and Grabe's analysis predicts three different phrase boundaries, which are schematically shown in Figure 3.6. 13 Note that in case of a tonally unspecified boundary tonal spreading of the last tone occurs, which is shown by multiple association of the final tone in Figure 3.6b and c.

(a)

e m a i l i : ^Ezzuzutzzi ^ΕΖΖΙΠΏΠΏ I L*

I Η

I H%

I L*

\

Η

Η*

L

Figure 3.6: Three distinct heights of intonation phrase boundaries; each box represents a syllable and shading represents accentuation.

Contrary to Fery and Grabe, the GToBI system (e.g. Grice and Baumann, 2002) proposes, a high and a low boundary tone. Further, based on the assumption of phrase accents any boundary tone is complex in that it is a sequence of a phrase accent plus boundary tone. In total, four combinations of edge tones are assumed. If a phrase accent has the same tonal value as the intonation phrase boundary, both specifications axe collapsed into one single label, e.g. T—%, where Τ is any tone (cf. Grice and Baumann, 2002). (The notation of a T—% calls for the question which empirical justification a sequence of a low phrase accent and low boundary tone, L—%, really has, or, in other words, whether such a tonal sequence might be better accounted for assuming tonal spreading of a trailing tone, as Fery (1993) and Grabe (1998a) do; yet, this is not our concern here, and we leave this interesting aspect of German intonation here as it is.) For the present study, we will assume a system based on Fery and Grabe. Tonal spreading of trailing tones yields in tonally unspecified boundaries. Representational aspects The previous parts of this section have drawn a line between the two distinct accounts of German intonation by means of tonal differences. This part here is concerned with an additional level, namely differences in tonal representation. In the origi13

The three-way phrase boundary system (cf. Figure 3.6) seems to be experimentally proven by Schneider and Lintfert (2003).

33 nal tone-sequence-model (Pierrehumbert, 1980) as well as in further developments of this model, e.g. Pierrehumbert and Beckman (1988) and American English ToBI (Silverman et al., 1992; Beckman and Ayers-Elam, 1997), tones have been treated as phonological entities on the surface that are transfered into actual pitch values by means of phonetic rules (cf. chapter 1, section 1.3). Alternatively, Gussenhoven (1984) has proposed a model of tonal representation where phonological entities, the pitch accents, are represented at an underlying tonal level, and surface representation is derived by tone linking rules as in (7). Pitch accents on the surface, then, undergo phonetic transformation as in the Pierrehumbert model. Gussenhoven's model is more restricted in that it assumes only bitonal pitch accents underlyingly. Any monotonal pitch accent is interpreted as a derived tone of an underlying bitonal accent category. The real advantage of the model, and the reason for us to assume this way of tonal representation adequate for our system of analysis, is its ability to account for the restricted distribution of monotonal pitch accents to prenuclear position. If one assumes only bitonal pitch accents, the tonal grammar is more economical, and monotonal accents are interpreted as surface variation of an underlying category. For German, this assumption is well evidenced by the fact that monotonal accents are restricted to prenuclear position, and thus derivable by tone linking rules (cf. Uhmann, 1991; Fery, 1993; Grice and Baumann, 2002, for the distribution of monotonal pitch accents in German). (7)

(a) partial linking and (b) complete linking (Gussenhoven, 1984) rp*rp

fp*^

'p+'p

^

y

T * >

'ρ*>

Τ

T * T

T*T

In (7-a), a rule for partial tone linking is given. In a sequence of two bitonal accents, the trailing tone of the first pitch accent is shifted to the right immediately prior to the second pitch accent, indicated by ' > ' . In (7-b), the rule for complete tonal linking is given where the trailing tone of the first pitch accent is deleted, resulting in a monotonal pitch accent. The rule of complete tonal linking thus accounts for monotonal pitch accents on the surface. Consider, for instance, a German example from Fery (1993), adapted in (8). A completely linked prenuclear accent results in a new phrasal structure of a sentence. In (8-a), two phrases are indicated by bracketing. In (8-b), the two accents are associated with only one phrase, i.e. "[ MANN mit dem MOTORRAD ]". (8)

Partial (8-a) and complete (8-b) linking of two falling pitch accents, from Fery (1993, 125-126). a.

H*

L

H*

L

( Lena verfolgt den) [ MANN mit dem MO] [ T O R R A D ] b.

Η*

Η*

L

( Lena verfolgt den) [ MANN mit dem MOTORRAD ] "Lena is following the man with the motorbike."

34

Summary Our discussion of German intonation has been concerned with drawing a line between two distinct approaches of German intonation. For a variety of reasons, we base our system analysing and comparing dialect intonation on the intonation model that is based on Gussenhoven (1984) and which Fery (1993) and Grabe (1998a) adapted to German. For a summary of our analysis system see section 3.4 below.

3.2.3

Downstep

This section intends to summarize briefly the notation of downstep in intonational phonology. Our analysis of SwabG and SaxG shows that downstep regularly occurs. In the Pierrehumbert (1980) model, tonal targets are evaluated by context-sensitive rules in order to account for downstep. The downstepped contours shown in Pierrehumbert (1980, Figures 4.1 to 4.4) decrease faster than one would expect if drawing a declination line. Thus, downstep is a characteristic intonation pattern for English, and also for other Germanic languages such as German and Dutch (Fery and Kügler, submitted; van den Berg et al., 1992). In her model, Pierrehumbert assumes that bitonal pitch accents trigger downstep, more precisely, that the trailing or leading tone of a bitonal accent causes downstep. If a certain tonal configuration occurs, downstep occurs. For falling contours a minimal pair like in (9) exists. (9)

A sequence of H* pitch accents as a minimal pair of non-downstepped (9-a) and downstepped accents (9-b) according to Pierrehumbert (1980). a. Η* Η* H* L - L% b. H * L - H * L - H* L - L%

A drawback of a notation like (9) that has been noted (see among others Ladd, 1983) is that one has to account for the final nuclear fall in pitch. Inspired by Bruce (1977), Pierrehumbert borrows the concept of the phrase tone to account for the pitch curve between the nuclear accent and the boundary. 14 Since the bitonal H*L— accent is occupied by its function to trigger downstep, Pierrehumbert chooses the theoretical solution of the phrase accent. Contrary to Pierrehumbert, Ladd (1996) and in particular Ladd (1983) assume that downstep is phonologically independent. While Pierrehumbert assumes the bitonal accents to trigger downstep, Ladd proposes an independent feature that might be combined with any phonological tone. Downstep is usually indicated by an exclamationmark '!'. Hence the contours in (10) as compared to (9).

14

Note that the phrase tone in Swedish is a tonal marker for focus, thus being functionally equivalent to a pitch accent in a language like English. It is its structural property being realised after the word accent fall and before an intonation phrase boundary that Pierrehumbert borrowed to account for the course of pitch between the nuclear accent and the intonation phrase boundary.

35 (10)

A sequence of H* pitch accents as a minimal pair of non-downstepped (10-a) and downstepped accents (10-b) according to Ladd (1996). a. b.

H* LH* H*L L% H* L!H* !H*L L%

Following this assumption, the individual tones (bitonal or monotonal) keep their tonal characteristics; a falling pitch accent of the form H*L remains a fall whether downstepped (!H*L) or not (H*L). Prom (10) it becomes clear that "downstep is an independent linguistic choice with identifiable meaning and identifiable phonetic effects" (Ladd, 1996, 92). What makes (10-b) distinct from (10-a) is the added feature of downstep that phonetically lowers the second and third accent in comparison to the respective preceding one. This phonetic distinction is additionally accompanied with a distinction in meaning, where "downstepping adds a nuance of greater finality but does not otherwise seem to affect the meaning of the contour" (Ladd, 1996, 76). Fery (1993, 158) notes for downstep in German: "if [...] two accents have the same prominence, then they are realised as a sequence of downstepped accents, and each of them forms an intermediate phrase. This downstep is larger than the effect of declination would lead us to expect." Accordingly, downstep is an indicator of prominence relations of two or more accents within an intonation phrase. Additionally, an intonation phrase in German with downstepped accents in contrast to non-downstepped accents yields different meanings that are closely connected to the focus structure of the phrase. "If it is downstepped, the whole phrase is understood as focused. If, on the other hand, the second accent is not downstepped, then the first part of the sentence is understood as being already given and part of the background." (Fery, 1993, 159). Fery illustrates this with two contexts that yield two distinct intonation patterns of the sentence Am meisten gegessen hat Rudi 'Rudi ate the most', cf. (11). (11)

A context that induces downstep on the target sentence (a), and one that does not trigger downstep (b), from Fery (1993, 159-160). a.

b.

(Wie war Toms Kindergeburtstag?) 'How was Tom's birthday party?' Es war schön, es gab viele Spiele und viele Kuchen, und, wie immer, am meisten gegessen hat Rudi. 'It was nice, there were games and many cakes, and, as usual, Rudi ate the most.' (Haben viele Kinder gleich viel gegessen?) 'Did many children eat the same amount?' Nein, am meisten gegessen hat Rudi, und am wenigsten Jürgen. 'No, Rudi ate the most, and Jürgen the least.'

In (11-a), the whole sentence is focused since the question does not focus a particular part of the target sentence. This results in a downstepped falling nuclear accent. The question in (11-b), on the other hand, specifically asks for the activity of eating. Therefore, the first part of the target sentence is already given, and the focus of the sentence is on Rudi, which is intonationally marked by a non-downstepped falling nuclear accent. Thus, tonal distinctions like in (10) occur in German as well, the different meaning is attributed to the focus structure of the sentence. These contours are thus not

36 only options that speakers may have (as it appears to be in English), but express clear differences in meaning in German. Another distinction of downstep between English and German is its distribution of occurrence. While a downstepped contour in German is perceived as the most natural intonation pattern of a declarative, in English it is not. In English, a more natural pattern for a declarative would be a non-downstepped falling nuclear accent (cf. Pierrehumbert, 1980). Although theoretically well studied, the phenomenon of downstep raises questions as yet unanswered. There are, to our knowledge, only two studies calculating the amount of downstep quantitatively, and this is done for American English (Liberman and Pierrehumbert, 1984) and for Dutch (van den Berg et al., 1992). In their experimental based model, van den Berg et al. propose an invariant downstep factor by which speakers lower their pitch range in relation to the preceding accent. It is thus a local factor. Additionally, they also propose a phrasal downstep factor, which is invariant as well. This means that even intonation phrases compared to each other may be lowered by a constant factor, however the inter-phrasal factor is claimed to be lower than the intra-phrasal factor. In an experimental investigation, Grabe (1998a, 185ff.) has found that German speakers have different options to implement downstep. The peak of the last pitch accent may be realised on a continuant scale between the pitch height of the previous peak and the phrase-final low pitch. An analysis of the standard deviation of the final pitch peak revealed this result, where mean peak height is significantly lower than the preceding peak. Grabe concludes that downstep in German may be partial or complete. Since this is a gradient implementation, Grabe rejects an analysis of a nuclear H + L * accent, where a pitch peak is absent on the accented syllable - a pitch accent that is included in the GToBI system (see contour 6b in Table 3.1 above) (Grabe, 1998a, 210). Recent studies have attempted to model the opposite phenomenon, i.e. upstep or the interruption of downstep (Truckenbrodt, 2002; Fery and Truckenbrodt, 2005; Truckenbrodt and Fery, Ms; Fery and Kügler, submitted). However, for German, and for many other languages, it is not clear where in the signal to draw the categorical line between a downstepped and a non-downstepped pitch accent. It is not certain, whether downstepped and non-downstepped contours really represent two distinct categories or whether there is a gradient phonetic scale. For our purposes, we assume downstep to occur regularly in our data, and most of the time, we identify a downstepped accent on perceptual grounds.

3.2.4

Linguistic function of intonation

Intonation usually functions to structure information in an utterance that a speaker wants to communicate to a hearer. Presumably, this is an uncontroversial and widely accepted function of intonation. Controversially discussed, however, are the conditions that render individual accent structures, and the interpretation of these accent structures in terms of meaningful categories. From a syntactic point of view, researchers claim that the syntactic structure of a sentence governs accent placement, an assumption which Chomsky and Halle (1968) have formulated as the nuclear stress

37 rule. Bolinger, in contrast, has argued that an accent expresses highlighted information rather than that an accent is governed by structural means (in particular Bolinger, 1972). Regardless of the drawbacks or advantages of these views, we will follow Ladd (1980, 76) assuming a weak view of Bolinger's theory which does not completely neglect the role of syntax: "accent goes on the point of information focus, unless the focus is unmarked, in which case the accent goes in a location determined by the syntax". For Ladd, this is the basis for distinguishing between broad and narrow focus, where broad focus is any kind of unspecified focus and represents normal or default stress, 15 and narrow focus comprises a certain, smaller constituent: "... accent - in general - goes on the rightmost accentable item of the focus constituent. If the focus constituent is the whole sentence, we get 'normal stress'; if not, we get a narrow focus on the constituent identified by the placement of the accent." (Ladd, 1980, 77, emphasis Ladd's).

The formation of focus domains is generally determined by contextual factors. Given these, Gussenhoven (1984) has proposed "Sentence Accent Assignment Rules" that account for stress placement. Given a certain accent structure, Selkirk (1995) has proposed focus projection rules. Since our investigation does not aim at contributing to the field of focus domain formation we will not further explore this issue here. We simply assume similar rules to apply in German dialects. For Standard German, Uhmann (1991), and Fery (1993) have shown this in detail. See also Fery and Herbst (2004) for a recent experimental investigation on sentence accent assignment rules in German. Having drawn a distinction between narrow and broad focus, sentences appear not only to consist of focused items, at least if narrowly focused. A common partition of a sentence refers to the Focus-Background structure. 16 Given the Focus-Background structure, Biiring (1996) integrates the 'topic' as being part of the 'non-Focus' partition, a view we will adopt here. A sentence structure according to Biiring is given in (12) where he distinguishes between a Focus and a non-Focus part that further branches into a Topic and a Background. (12)

Structure of a sentence according to Biiring (1996, 47). S Focus

non-Focus Topic

15

16

Background

"Focus can apply to constituents of any size. Normal stress is simply the accent placement that permits the broadest possible focus interpretation - focus on the whole sentence" (Ladd, 1980, 78). We are aware of several other proposals referring to information structural parts of sentences. Among others, partitions of sentences have been called thema-rhema, topiccomment, or given-new. The reader is referred to Fery (1993) for an overview.

38 The obligatory part of a sentence appears to be the Focus, i.e. a case of an 'out-ofthe-blue-sentence' or a so-called 'citation form' - broad focus (13-a). Hence, either the Background (13-a), (13-b), the Topic (13-a) and (13-c), or both (13-a) may not be part of a sentence. In (13-d), Maria is the topic, fährt background, and nach Berlin focus. (13)

Different contours signaling different functions, adapted from Fery (1992, 2ff); topic and focus marking according to Büring (1996). a.

b.

c.

d.

Irgendwas neues? 'Anything new?' [ Maria ist nach Berlin gefahren. ]p0c H*L !H*L Übrigens Maria, was hat sie gemacht? 'What about Maria, [ Maria ]χ 0 Ρ [ ist nach Berlin gefahren. ]FOC what did she do?' L*H H*L Wer ist nach Berlin gefahren? 'Who went to Berlin?' [ Maria ]p0c ist nach Berlin gefahren. 'Maria went to Berlin.' H*L Wohin fährt Maria? 'Where is Maria traveling to?' [ Maria ]τ 0 ρ fährt [ nach Berlin. ] f o c 'Maria is traveling to Berlin.' H*L H*L

A topic is a constituent which the speaker is going to say something about (the theme of the sentence). 17 A topic may be given by the context, for instance erased by means of a question, e.g. (13-b) and (13-d). If Maria were unaccented in (13-d), it would belong to the background signaling no special emphasize of this constituent. The accentuation of a given constituent is thus optional (cf. also Fery, 1993). Besides the topic as a (contextually) given constituent, Büring discusses three specific kinds of topics, i.e. the contrastive, the partial, and the purely implicational topic, which are different uses of topics without any semantic difference (Büring, 1996, 50). A contrastive topic moves "the conversation away from an entity given in the previous discourse" (ibid., 49) as in (14-a). The topic Ί ' contrasts with 'Fritz' in the question. In (14-b), Β answers the question not exhaustively, thus the topic being only part of the answer. And in (14-c), the accent on the topic suggests a reading like A's wife might have kissed another man however. (14)

Different kinds of topic according to Büring (1996, 49, 50). a. b. c.

17

A: B: A: B: A: B:

Do you think that Fritz would buy this suit? Well [ I ] T certainly [ WOULDN'T ] F W h a t did the pop stars wear? The [ female ] τ pop stars wore [ caftans ]p. Did your wife kiss other men? [ My ]T wife [ didn't ]F kiss other men.

"It [the topic] is understood as 'what the rest of the sentence is about,' or 'the entity anchoring the sentence to the previous discourse'." (Büring, 1996, 48).

39 What is important to note here is that despite the obviously different kinds of topics, semantically as well as phonologically these are treated as a single instance of topic. The data in Fery (1993, 129ff) suggests that a topic is indicated by a rising pitch accent in German. A topic may also be a syntactically topicalized constituent as in (15), where GESCHLAFEN is topicalized. 18 Again, the topic here is intonationally marked by a rising accent, indicated by a slash ' / ' in (15). (15)

[IP [ ip GESCHLAFEN /] [ ip hat KEINER \ von uns ]] 'None of us slept.' (Fery, 1993, 129)

From a phonological point of view, focus as well as topic are signaled by an accent. Languages and/or dialects differ in the way which kind of accent expresses which kind of focus or topic. In Standard German, for instance, a nuclear falling accent expresses broad and narrow focus (Fery, 1993).19 Other languages, such as European Portuguese (Frota, 2000), Russian (Alter, 1997a,b; Zybatov and Mehlhorn, 2000), or Bengali (Hayes and Lahiri, 1991), have different accents to express broad and narrow focus. Recently, Peters (2001) has shown for Berlin German, that rising-falling accents express contrastive focus. For topic, Standard German - as well as Swabian and Upper Saxon German studied here - mark a topic by means of a rising accent (cf. Fery, 1993, and chapters 4 and 5). In English, however, a topic appears to be marked by a rise-fall (accent Β according to Bolinger, 1958) as in (16), the famous example of Jackendoff (1972) (see Fery, 1992, 1993, for a thorough review of English rise-falls and the relation to German rising accents). (16)

A: B:

Well, what about FRED? What did HE eat? FRED ate the BEANS

(Jackendoff, 1972, 261)

However, recent work on the phonology of information structure suggests that there is no one-to-one correspondence between a certain type of pitch accent and an information structural status Fery (2007). A further dimension of information structure concerns givenness, a factor which often has been neglected. The level of givenness seems to be orthogonal to concepts such as topic-focus since any part of the sentence, i.e. topic, background or focus, may or may not be given (for an overview see Baumann, 2006). According to Baumann, givenness contrasts with accessible and new information, and the different cognitive states of discourse referents are expressed by different linguistic means. In particular, Baumann proposes a model of intonation and givenness where newness correlates with a high pitch accent. At the other end of this gradual scale of givenness, completely activated information correlates with deaccentuation. In between, accessible information seems to be associated with a falling accent H+L* (Baumann, 2006). 18

19

.Jacobs (1982) called this kind of topicalization I-Topikikalisierung Ί-topicalization', where Ί' means indicated by intonation. Note that GToBI assumes two different accents to signal broad and narrow focus: a falling contour (H* L-) for broad focus, and a rising-falling contour (L+H* L-) for narrow focus (e.g. Grice and Baumann, 2002). See also Peters (2006) for a similar proposal.

40 3.2.5

Intonation and Meaning

In the preceding section, we assumed that intonation usually functions to structure information in an utterance that a speaker wants to communicate a hearer. With respect to intonational meaning, Gussenhoven (1984) has proposed a theory which assumes three distinct meanings that intonation may express. For English intonation Gussenhoven has shown that each of the three meanings can be ascribed to a distinct nuclear contour. Carrying over to our analysis we attempt to apply Gussenhoven's model to SwabG and SaxG intonation (cf. chapter 7 where we will contrast Gussenhoven's model with a different approach to intonational meaning put forward by Pierrehumbert and Hirschberg, 1990). We briefly introduce core aspects of Gussenhoven (1984) here. In his theory of intonational meaning, Gussenhoven (1984) anchors the communicative functions of intonation. In a communicative situation, speakers intend to "strive towards some common understanding about a particular segment of the world. The understanding they think they have reached at any one point in that process [communication, F.K.] is called the background" (Gussenhoven, 1984, 201). According to Gussenhoven, a speaker modifies or refers to the background having three basic options: a speaker may add or select a variable to / from the background, or a speaker "may choose not to commit himself as to whether the Variable belongs to the background", i.e. test the variable (Gussenhoven, 1984, 201). Further, Gussenhoven assumes that these three options are expressed intonationally. He considers nuclear tones, i.e. nuclear contours, to convey intonational meaning, thus being morphemes. For British English, Gussenhoven assumes three morphemes to be basic, the fall (H*L), the fall-rise (H*LH) and the rise (L*H), which convey the meaning of A D D I T I O N , SELECTION and RELEVANCE TESTING, respectively (cf. also Gussenhoven, 2004, 296ff). To account for the intonational diversity of English intonation, Gussenhoven assumes modifications of the base contours. Modified contours are derived from the basic ones and share structural as well as functional similarities with the corresponding base contour. Gussenhoven assumes that the tone that erases by means of a modification is a morpheme on its own, yet it depends on the base contour. The crucial assumption is that a modification may affect every base contour, adding the particular meaning of the modification proper to the meaning of the base contour. 20 Thus, the tonal and the semantic properties of a modification are constant across contours. The three base contours are U N M O D I F I E D , whereas Gussenhoven assumes three modifications to affect the unmodified ones, DELAY, STYLISATION and H A L F - C O M P L E T I O N (cf. Gussenhoven, 1984, 215ff). According to our analysis of intonational meaning in SwabG and SaxG proposed in chapter 7, we assume, for the time being, only one modification to apply in the two dialects, namely L-affixation which expresses the same kind of meaning as 21 DELAY in English. 20

21

Following Gussenhoven ( 1 9 8 4 , 2004) we assume that a modification affects the whole nuclear contour and not only pitch accents as argued by Ladd ( 1 9 8 3 ) . Note that Gussenhoven ( 2 0 0 4 , 307) has changed the label of the morpheme DELAY to L-prefixation, which goes back to Gussenhoven and Rietveld ( 1 9 9 2 ) .

41 3.3

Accounts of dialect intonation

In recent years, intonational variation has received growing attention (for a current overview see Gilles and Peters, 2004b). In a large-scale research project, the intonation of urban varieties of Germany is investigated (cf. Auer et al., 2000; Gilles, 2005; Peters, 2006). The main goal of that project is to analyse and describe regionally salient intonation contours both phonetically and phonologically. A similar project on English has focused on phonological and phonetic realisational aspects of intonation in several dialects of British English (Grabe and Nolan, 1997-2002; Grabe, 2004). Dialectal variation in Swedish concerns the documentation of Swedish varieties spoken around the year 2000 (e.g. Engstrand et al., 1997; Bruce et al., 1999), where intonational aspects touch upon the relationship between word accent and phrasal intonation in the dialects (Bruce and Thelander, 2001), as was done for Standard Swedish (Bruce, 1977). Our study of SwabG and SaxG intonation contributes to the field of regional variation supplementing the studies of Gilles (2005) and Peters (2006). According to Ladd (1996, 119), the intonation of languages may differ on four distinct levels, i.e. semantically, systemically, realisationally, and phonotactically. 22 We assume that Ladd's taxonomy may hold for dialectal variation as well. Our analysis of SwabG and SaxG intonational patterns in chapter 4 and 5 reveals differences in the tonal inventory of these dialects. In addition, similar contours are shown to differ in phonetic realisation in chapter 6. The following sections outline some of the comparative approaches to dialect intonation with respect to realisational and systemic differences in intonation.

3.3.1

Realisational differences

Realisational differences in intonation are similar to segmental variation between dialects. Considering VOT, for instance, dialects differ in the amount of VOT across a given category (for an overview see Braun, 1996). With respect to intonation, realisational aspects concern the timing and excursion of accentual peaks and valleys, accent alignment, and phonetic adjustments under time pressure (truncation vs. compression). For the study of comparative dialect intonation, these features become highly relevant, as Gilles (2005) has noticed. Concerning the timing of peaks, Bruce and Gärding (1978) and more recently Bruce and Thelander (2001) have presented a typology of dialect differences of the word accent distinction in Swedish. Based on the work of Bruce (1977) for Standard Swedish, where Bruce identified the phonetic correlates of the word accent distinction in Swedish in addition to phonetic correlates of focus and boundary specifications, Bruce and Gärding (1978) analysed the timing of the pitch accent gesture in different Swedish dialects. The general distinction between accent I and II in Standard Swedish 22

Fitzpatrick (2000) provides a similar classification, uses however the terms "phonetic" and "phonological" instead of "realisational" and "phonotactic".

42

(Bruce, 1977) has been found in dialects as well: the timing of the word accent fall is earlier in accent I than in accent II, but Bruce and Gärding (1978) established a typology of dialects where they correlate the timing of the pitch accent gesture from early to late with the geographical distribution of the dialects from east, west, south, and central Swedish. Bruce and Thelander (2001) identify the transitions between the southern and western dialect border in terms of the accent typology of Bruce and Gärding (1978). Regarding accent alignment, Atterer and Ladd (2004) have shown that accentual targets of prenuclear rising accents align fairly stable with the segmental string. The alignment pattern within a dialect is invariant but differs between dialects if comparing identical accents. Accordingly, the alignment of accentual tones seems to be a language- and dialect-specific phonetic manifestation. Atterer and Ladd found that Southern German speakers align the accentual gesture of a prenuclear rise significantly later than Northern German speakers do. With respect to phonetic adjustments under time pressure, Erikson and Alstermark (1972) and Bannert and Bredvad- Jensen (1975) have shown that, if less sonorant material in an accented syllable is available, languages may have the option of cutting the accentual gesture off at the transition of voiced to unvoiced segments (i.e. truncation), or to compress the whole accentual gesture onto the minimized string of voiced segments. Bannert and Bredvad-Jensen (1977) have applied this method to highlight dialectal differences with respect to accent realisation. More recently, Grabe (1998b) and Grabe et al. (2000) have shown that measuring truncation or compression yields in language and dialect specific characteristics where Standard German appears to truncate falling accents, while Southern British English, compress (Grabe, 1998a,b). Comparing this result for English with several varieties of English spoken in the British Isles, Grabe et al. (2000) found that dialects of a language differ considerably in terms of phonetic adjustment of pitch accents: while Standard British English as well as most of its varieties compress falling accents, Belfast and Leeds English truncate falling accents (Grabe et al., 2000; Grabe, 2004). Similarily, Standard German as well as most varieties of German truncate falling accents (Grabe, 1998a,b), except the Northern variety of Hamburg (Gilles, 2001a; Peters, 1999b) and the North-Western variety of Cologne 23 (Gilles, 2005). Kügler (2003b) showed that the concept of truncation and compression can be extended to syllable structures containing more sonorant material than only the syllable nucleus (see also Peters, 1999b). Using this method, Kügler (2004b) has provided evidence for compression in Stuttgart Swabian German nuclear rising accents. Nuclear falling accents do not belong to the tonal grammar of that dialect (cf. chapter 4). The most comprehensive study of realisational aspects of German dialect intonation is the study by Gilles (2005). The investigation is based on two functionally distinct categories of analysis, the suspensive and the final phrase (cf. Klinghardt, 1925; Barker, 1925, for this classification). These units are identified independently of prosody. Gilles compares the intonational patterns realised in these categories between eight different urban varieties of German. 24 23 24

Compression has been shown only for accent 1 words in Cologne (Gilles, 2005, 215ff). The urban varieties analysed in Gilles (2005) are Berlin, Dresden, Duisburg, Hamburg, Cologne, Mannheim, Munich, and Freiburg.

43 For final phrases, Gilles (2005) observes mainly falling contours, with the exception of Freiburg. (Recall that Freiburg German belongs to the same dialect area as SwabG, cf. chapter 2.) Falling contours vary considerably with respect to the position of the peak and the shape of the fall. The position of the peak in the accented syllable depends on the dialect. It is realised earlier in western German varieties (e.g. Hamburg, Cologne, Duisburg and Mannheim) and later in eastern German varieties (e.g. Berlin, Dresden and Munich). The shape of the fall is very fast initially, and slows down after a mathematical turning point in the curve (cf. the term "Knickkontour" which Gilles, 2005, uses to express this point). The dialects vary with respect to the turning point, the earlier the peak is realised in the accented syllable, the longer is the falling shape and the later is the turning point in pitch. A variant of the falling contour lacks the low trailing tone according to Gilles' analysis (H* L%). The contextual distribution of this contour appears to announce a longer turn, e.g. in case of a narrative. This expression of cohesion coincides with completely linked prenuclear accents that re-phrase an utterance, thus expressing a cohesion between the two phrases. This similarity in intonational function appears to be further evidence to justify tone-linking rules in nuclear position as well (see (7) above). In Freiburg, as well as in Dresden and Mannheim, a dialect specific contour occurs in signaling finality. In Freiburg, it is a rising falling contour that signals finality, no simple falling contours are claimed to exist (cf. also Peters, 2006), which is similar in SwabG as our analysis will show (cf. chapter 4). For Dresden, in addition to the falling contours, an extended falling curve "Fallbogen" (Gericke, 1963) occurs signaling finality (for more about the Fallbogen see chapter 5). For suspensive phrases, Gilles (2005) analyses more contours than for final phrases. All of these contours share the characteristic of ending high phrase-finally. As for the simple falling contours, we would suggest modeling some of these contours as being the result of tonally linked contours. In total, however, Gilles has found more variation with respect to suspensive phrases than with respect to final ones. In sum, we have shown that dialects vary with respect to the phonetic realisation of tones. Parameters such as peak timing, alignment and truncation /compression significantly characterize a language variety, which suggest that phonetic aspects may be captured assuming language- or dialect-specific implementation rules (Atterer and Ladd, 2004).

3.3.2

Systemic differences

In terms of cross-language comparisons, systemic differences concern the "inventory of phonologically distinct tune types" (Ladd, 1996, 119). Given two languages with their respective tonal inventories, a comparison of these is 'systemic' (cf. Grabe, 1998a, who compares English and German). The most comprehensive study on dialect variation with respect to systemic differences of regional varieties of German is Peters (2006).

44 In his comparison of six urban varieties of German, 25 Peters proposes a theory of "phonological equivalence" in order to compare the individual tonal contours between the dialects phonologically with simultaneous reference to the phonological status of any individual contour within the phonological system of its language variety. Peters displays the intonational system of each dialect in a finite-state grammar. For comparative purposes, however, Peters assumes an abstract phonological level based on tonal features such as [± high], [± late peak], [± falling], or [± left-headed]. An inventory of a dialect is described on the basis of tonal features. Since tonal labels carry redundant information with respect to tonal features, some of the labels receive an abstract notation Τ (for tone) whose actual value is inferable from the features (cf. (17)). According to Peters' phonetic implementation rules, the abstract tonal marker receives a phonetic value. In (17), implementation rule (17-b) applies. (17)

a. b.

H*T, with a feature [+ falling] H*L Application of Rule A: A trailing tone becomes low after a high, and high after a low accentual tone (Peters, 2006).

The abstract representation allows for a comparison of contours that appears to be equivalent. Consider for instance a simple rising contour, which can be expressed as in (18). This contour has been analysed for Standard German, Hamburg, Berlin, Duisburg, and Freiburg (cf. Peters, 2006).26 (18)

A simple rise in five varieties of German: L*H 0l (l refers to an intonation phrase boundary (%), cf. Gussenhoven, 2004).

At first glance, these contours appear to be equivalent since the similarity in tonal labelling suggests a similarity in phonological form. The similarity, however, refers only to a rough phonetic notation where not all phonetic features are represented. Apart from this, Peters assumes that a comparison of the individual tones does not refer to their phonological status within the inventory of the respective variety. This becomes obvious in (19) comparing this contour with a similar one that occurs in Freiburg. (19)

A simple rise in Freiburg German (according to Peters, 2006) a. in terms of a rough phonetic notation: L*H 0l (cf. (18)) b. in terms of a phonological notation: T*T &l

In Standard German, as well as in Hamburg, Berlin and Duisburg, L* and H* express a contrast in the phonological system. Therefore, the feature [+ low] is a distinctive feature. Concerning Freiburg German, however, Peters assumes that L* does not contrast with H* since no categorically distinct high accents occur in that variety. 25

26

The six varieties analysed by Peters (2006) are Hamburg, Berlin, Duisburg, Cologne, Mannheim, and Freiburg. Peters (2006) analyses a similar contour in Cologne and Mannheim as well. However, in Cologne and Mannheim this contour exhibits an additional low phrase boundary tone. We exclude these two from our review, since we simply attempt to show Peters' theory of phonological equivalence with surfacely similar contours.

45 The low accentual tone L* bears the feature [+ low], yet it is only a phonetic feature since it is not distinctive phonologically. According to Peters, the nuclear rising accent differs between Standard German and Freiburg German with respect to its phonological content, since the phonological content is defined on the basis of the number of distinctive features. This difference, Peters expresses by means of abstract phonological notations. In sum, the comparative system proposed by Peters (2006) allows for two comparisons at the same time. Cross-dialectally, it allows for a comparison of phonetically equivalent contours, thus a kind of syntagmatic comparison. At the same time, the abstract tonal notation of a contour gives reference to its phonological status within the phonological system of the language variety, thus it is a kind of paradigmatic expression.

3.4

An approach of modeling dialect intonation

This section intends to propose an autosegmental-metrical system that allows for cross-dialect analysis and comparison. The system is characterized by the following criteria: - Two levels of tonal representation - Underlyingly, only complex pitch accents are assumed - Intonation phrase boundaries may be left tonally unspecified - Nuclear contours are considered to be morphemes - Phrase accents are rejected for analysis - Intonational variation has to be comparatively modeled in terms of realisational, phonotactic, semantic, and systemic differences Regarding the levels of tonal representation, we follow the two-levels-of-representation approach (Gussenhoven, 1984) here for two reasons. First, this model seems to be more economical with respect to the assumed tonal grammar. Second, this model seems to be appropriate since it more adequately captures tonal variation on the surface assuming tone-linking rules (cf. (7) above). We further assume that the tonal morphemes underlyingly are complex. In a complex accent, one tone always is more prominent than others (cf. Pierrehumbert, 1980). According to the theory, bitonal accents may have either a less prominent leading or trailing tone, in combination with a more prominent accentual tone (cf. (3) above). On the surface, however, monotonal pitch accents may occur, and are analysed as a result of a completely linked underlying bitonal pitch accent (cf. (7-b) above). Given our discussion on boundary tones above, we concur with Fery (1993) and Grabe (1998a) that intonation phrase boundaries may be left tonally unspecified. If left unspecified, a boundary receives a tonal value by spreading of the trailing tone of a nuclear pitch accent (cf. (5-a) above), forming either a low or high plateau. Following Gussenhoven (1984, 2004), we assume nuclear contours to be morphemes. That is, a nuclear contour as a whole conveys a certain kind of meaning. We further

46 assume that Gussenhoven's model of intonational meaning proposed for English can be carried over to cover SwabG and SaxG intonation. In line with Gussenhoven (1984), we assume that three basic types of meanings can be distinguished, i.e. A D D I T I O N , S E L E C T I O N , and R E L E V A N C E T E S T I N G . Our analysis of intonational meaning reveals that the two dialects express these meanings intonationally by means of different contours (cf. chapter 7). Finally, the cross-dialect comparison of intonation patterns has to be done on different levels. We assume that the proposed taxonomy of language comparison by Ladd (1996, 119) also holds for purposes of dialect comparisons. In particular, Gilles (2005) and Peters (2006) have shown that dialects can be compared on the basis of 'realisational' (Gilles, 2005), and 'systemic' (Peters, 2006) differences. Our analysis of SwabG and SaxG intonation patterns is based on these criteria.

4

Swabian German Intonation Patterns

„Einen abweichenden Tonfall haben die schwäbisch-alsässisch-schweizerischen Mundarten. Hier hat die starkbetonte Silbe einen tieferen Ton als die neben- und schwachbetonte Silbe. Der abschwellenden Wortbetonung entspricht ein steigender Worttonfall." (Bremer, 1893, 195) 1

According to Otto Bremer's observation at the end of the 19 i/l century, Southern German intonation appears to deviate from - at least - Northern German intonation. Two decades later, Sievers (1912) hypothesized two basically inverted intonation systems in German. He claims that Southern German intonation mirrors that of Northern German. Given high or falling accents as the neutral accentuation in Standard German (e.g. among others Fery, 1993; Grabe, 1998a; Grice and Baumann, 2002), Southern German including the SwabG dialect is supposed to have low or rising accents. Studying the impressionistic work of dialectologists tends to support such an impression. The Southern German variety of SwabG has received little attention, except in Frey (1975). As it happens that certain intonational features are best characterized as being Southern German rather than dialect specific, we extend the following review of previous dialect intonation research to Southern German intonation (section 4.1). This allows for comparisons between related dialects and certainly helps understanding the peculiar position of SwabG intonation within the German speaking area, since, as Bremer already has pointed out, at least the South-Western varieties of German appear to diverge from Standard German intonationally. The main part of this chapter presents our analysis of SwabG intonation patterns, and the main focus is on nuclear contours (section 4.2) where we distinguish three base contours and a modified contour. Aspects of prenuclear accents are discussed in section 4.3. The chapter closes with a discussion about SwabG dialect specific intonation features (section 4.4).

4.1

Southern German Intonation

In his description of his own Stuttgart Swabian dialect Frey (1975) tries to explain the singing quality of Swabian intonationally. He gives evidence for gliding tones („Gleittöne", p. 67), the most common pattern of that dialect. These gliding tones correspond to Bremer's observation of low pitched, stressed syllables. Frey analyses these as a phonemic sequence shown in (1-b) following the American structuralist tradition of intonational description (e.g. Pike, 1945). Frey's description of intonation is based on four tone levels, 1 being the lowest and 4 the highest level. Additional 1

"The Swabian-Alsatian-Swiss dialects have a differing accent. Here, the strongly stressed syllable has a lower tone than the level or weakly stressed syllable. A rising word accent corresponds to the reducing word stress." (Our translation of Bremer, 1893, 195)

48 symbols are used to indicate secondary information of tonal levels: the degree symbol indicates sentence stress („Satzton tragende Silbe", p. 67), and a downward arrow indicates falling intonation at the end of a sentence („fallende Endkontur", p. 68). Prey illustrates this intonation pattern with a question-answer pair that is reproduced here in (1), see also the original in Figure 4.1. (1)

a. b. c.

/,wan koms 'den/ /,am 'naxmi'däg/ °13 1* °1 3 1 1 L * H L% L * H L% 'When will you come? - In the afternoon.'

;- /A /,Kan "'komS ' d e n / / , am 1. Wann kommst du denn? a u d i gewiß?

/

- O β'naxmi'däg/

Ζ.

(our own tonal adaptation)

-- - /

/'komS.ao Am N a c h m i t t a g . 3.

"gvis/ Kommst du

//

/ ' h a e 1 f r a i l a / / , v e n s " r t g a n d | 'nemS , h a l d an " S i r m i. J a , f r e i l i c h . 5. Wenn es r e g n e t , nimmst du h a l t Schirm ( h e i m ) .

ν Γ ΛΜ' ('hoem)/ einen

Figure 4.1: Swabian German intonation, from Frey (1975, 68). If we compare the tonal realisation of the wh-question in (1-a) with the answer, i.e. a declarative sentence, we observe that the same tonal pattern is used in both sentence types (see (1-b)). According to Prey, the relative low main sentence stress (tonal level °1) followed by a sharp rise (tonal level 3) is responsible for the gliding perceptual quality of SwabG. Our interpretation of Prey's tonal analysis is given in (1-c) applying the framework of the tone-sequence model summarized in chapter 1, section 1.3. Our tonal adaptation in (1-c) reveals that wh-questions and declarative sentences in SwabG are realised with a rising nuclear pitch accent ( L * H ) followed by a low boundary tone (L%). (1-c) presents our analysis of SwabG intonation which we will present in the following sections. We analyse the default or neutral rising-falling accentuation as a sequence of a rising pitch accent and low boundary tone. A yes-noquestion however seems to be realised with a tonal pattern distinct from wh-questions and declaratives (cf. the last phrase in the upper panel of Figure 4.1). Prom the example given in (1) it is not clear whether the second level 1 tone, the one before the downward arrow, belongs to the pitch accent or forms a low boundary tone, as proposed by Frey. He points out that the tonal fall after the pitch peak does only occur if no further syllable with word stress occurs after the nuclear pitch accent. Frey takes the optional quality of the fall as evidence for a rising pitch accent with the fall being part of the boundary specification. Our analysis of SwabG intonation provides evidence for this claim, since we observe that the postaccentual fall always aligns with the intonation phrase boundary independent of whether the accent is realised early or late in the phrase (see section 4.2.2).

49 In her analysis of German intonation Fery proposes a tritonal accentual movement, i.e. a rise-fall (L*HL). In Standard German, Fery (1993, 94) restricts the function and use of the rise-fall pattern to an expression of "of course", and to television police inspectors addressing suspects. In addition, Fery points to the frequent use of the rise-fall in Southern German dialects, in particular in SwabG; an example is illustrated in Figure 4.2. Fery's analysis does not specifically point to dialectal variation but assumes this accent pattern to belong to the inventory of German intonation. In order to account for the rise-fall, Fery claims a tritonal pitch accent, a low accentual tone followed by a complex HL trailing tone. This assumption has been criticized (Gussenhoven, 1994; Ladd, 1994). However, given that Fery assumes no low boundary tone to exist in German - an assumption that, to our point of view, Fery has substantiated phonologically (see also Grabe, 1998a) - the complex trailing tone solves this issue.

Figure 4.2: Tritonal nuclear pitch contour in SwabG on the phrase Das war eine Biene 'That was a bee', from Fery (1993, 95).

Given the distribution of the rise-fall, one might wish to distinguish between a risefall in Standard German and a rise-fall in SwabG. Both contours share their general contour shape, i.e. rising-falling, but appear to have different origins phonologically. According to our analysis put forward in the next sections, we claim that the risefall represents the neutral intonation pattern of SwabG. According to Fery (1993), the neutral accent pattern in Standard German is a simple fall (H*L). Thus, from a functional perspective, the SwabG rise-fall resembles the Standard German simple fall functionally. The Standard German rise-fall, on the other hand, could be interpreted as a modified simple fall. Fery (1993, 93ff) argues that the rise-fall is similar to Ladd's feature [delayed peak] (Ladd, 1983). Gussenhoven (1984) proposes that a delayed peak contour has undergone the modification DELAY where the original contour - in the Standard German case a simple fall - is shifted rightwards, resulting in a low accented syllable. Yet, the meaning of the original contour remains, enriched with the additional meaning 'significant' that the morpheme DELAY conveys (Gussenhoven, 1984). The phonological form of a modified contour is different from its base contour; in the Standard German example discussed here, a rightwards shifted simple fall H*L would become a rise-fall L*HL. Presumably, this is what Fery (1993) expresses in (2-a). (2)

a. b.

L* HL Frau Kramer 'Mrs. Kramer' L+H*L Frau Kramer 'Mrs. Kramer'

Fery (1993, 94) our proposal

50 Inspired by our analysis of SaxG (cf. chapter 5) which is based on and itself inspired by Gussenhoven (2004), we might think of an alternative phonological structure of the Standard German rise-fall proposed in (2-b). Taking the idea of L-prefixation (Gussenhoven, 2004) seriously, in SaxG we assume a modified falling contour that has a low tone as a prefix. This prefix does not change the general accentual shape, metrical properties and prominence relations of the original contour. Therefore, we assume a low leading tone. If carrying this over to Standard German, a modified simple fall would result in a structure like (2-b). Of course, this has to be proven phonetically, both on acoustic and perceptual grounds, as well as phonologically with relation to the tonal inventory of Standard German. There are still good reasons to assume a tonal structure like (2-a) in Standard German. Whatever interpretation holds for Standard German, we still suggest a differentiation between a Standard German and a SwabG rise-fall, since the former is best analysed as a modified contour, while the latter represents a base contour of SwabG. In a broader geographical perspective, Gibbon (1998) makes the cursory observation that South Western German intonation in general is characterized by a rising pitch accent, labelled as L*+H. According to Gibbon the accentual peak is reached in one of the postnuclear syllables. Hence, accent realisation is characterized by a delayed pitch peak. Gibbon says that such an accent pattern is "common to a chain of dialects along the Rhine valley, from Switzerland ('Schwyzer Dütsch') to Cologne ('Kölsch')" (Gibbon, 1998, 93), a claim made without providing any evidence. For Alemannic German, a South-Western German dialect (cf. Wiesinger, 1983) and geographically closely related to Swabian German (cf. chapter 2, section 2.2.1), two studies have recently been published (Gilles, 2005; Peters, 2006). Both are based on an analysis of the urban variety of Freiburg. Gilles' analysis of urban varieties of German attempts to provide an overview of realisational differences between dialects (cf. chapter 3, section 3.3.1). He observes that nuclear rising pitch accents in Alemannic German appear to be the salient dialectal feature. According to Gilles, two distinct nuclear contours are involved, the rise-fall and the simple rise. The rise-fall is illustrated in Figure 4.3. While the rise-fall, labelled as L * + H HL- (L%), occurs in final intonation phrases, the simple rise, labelled as L * + H %, occurs in suspensive intonation phrases (Gilles, 2005). HL- L%

h- 1%

σ (σ' σ), franZO se drin



wa re

"Franzosen drin waren'

'einiges gebracht'

Figure 4.3: Nuclear rise-fall contours in Freiburg Alemannic German, from Gilles (2005, 327). In both cases, the accentual syllable is low, a following peak is reached in the following postaccentual syllable. In the left panel, the fall extends to two syllables, in the right panel to one syllable.

51

Gilles (2005) observes no falling contours which holds for SwabG as well (see our analysis below). It is noteworthy that Gilles introduces a bitonal phrase accent to account for a postnuclear high plateau that ends in a fall on the head of the phrasefinal foot (cf. Figure 4.3). In addition, Gilles observes that this contour occurs along the Rhine valley up to Cologne, i.e. in the urban varieties of Mannheim, Cologne and Duisburg, an observation corresponding to Gibbon (1998, see above). The second observed contour in Freiburg corresponds to a simple rise where the accented syllable exhibits a low tonal target with a subsequent rise to a to high plateau. Again, these contours have a distinct low tonal target in the accented syllable in common. The approach taken by Peters (2006) is a systemic one (cf. chapter 3, section 3.3.2) and attempts to cover the tonal grammar of Freiburg Alemannic German (as well as of five other urban varieties). According to Peters the tonal grammar of Freiburg Alemannic consists of three base contours, a rise-fall, a double rise, and a rise-plateau contour, cf. Figure 4.4.

Figure 4.4: Tonal grammar of Freiburg Alemannic German, from Peters (2006).

In line with Gilles (2005), Peters (2006) notes that the rise-fall contour is the most common (and most frequent) pattern of Freiburg Alemannic German, and it resembles the simple fall in Standard German in its function. Regarding the fall of the rise-fall, Peters acknowledges that the fall occurs in the head of the phrase-final foot (cf. Gilles, 2005), yet his analysis differs slightly from that of Gilles. Whereas Gilles introduces a complex phrase accent (HL—) to account for the postaccentual tonal shape, Peters (2006) argues for a complex phrase accent where the high part is analysed as the head of the complex falling phrase accent (indicated by the phrase index l, HtL). Peters analyses two further contours, the rise-fall-plateau contour (L*H Ht+L 0t) and the rise-fall-rise (L*H Hi+L Hi). According to Peters, both contours relate to the rise-fall where the former appears phonetically as an uncompleted final fall, the latter differs from the rise-fall in that it exhibits an additional final rise at the end of the phrase. Swiss German belongs also to the Southern German dialect area (cf. chapter 2, section 2.2.1; cf. Wiesinger, 1983), and rising-falling nuclear contours have been observed in Swiss German as well. Fitzpatrick-Cole (1999) examines the dialect of Bern Swiss German from a typological perspective, and claims phonological rather than phonetic differences to be the source of the nuclear accent difference between Bern Swiss German and Northern Standard German. 2 Accordingly, Fitzpatrick-Cole proposes a phonological structure of the rising nuclear accent as L*+H. Since the accent pattern in Bern Swiss German consists of a postaccentual fall, Fitzpatrick-Cole assumes a low boundary tone (L%), "which has the option of being stress-seeking" (1999, 944). 2

Note that the term 'phonological differences' used by Fitzpatrick-Cole (1999) and later (Fitzpatrick, 2000) corresponds to the term 'systemic', proposed by Ladd (1996).

52 Swiss German speakers

falling pitch patterns (HL)

rising pitch patterns (LH)

Figure 4.5: Schematical pitch contours in Standard Swiss German, from Ulbrich (2004, 113). In the left panel, two variants of falling accents are displayed, in the right panel two variants of rising accents.

This corresponds to the findings of Gilles (2005) and Peters (2006) who notice that the postnuclear fall is realised on the head of the phrase-final foot. Ulbrich (2004) compares phonetic cues of Standard Swiss German intonation with that of Standard (Northern) German analysing the speech of news readers. Ulbrich claims that news readers represent the respective standard variety, which she substantiates reporting on perception tests where linguistically trained and naive listeners allocated the speakers' origin almost a 100 % correctly and rated them as being speakers of the respective standard variety. Acoustically, Ulbrich reports on measurements of speech rate, pitch range, and pitch accent realisation. Apart from significantly longer accented syllables for Swiss German speakers (a fact that is identical to our analysis of the duration of SwabG accented syllables as compared to SaxG ones, cf. chapter 6, page 148) and a slower speech rate, most interestingly, Swiss German speakers use a rise-fall (LH_L(%)) 3 more often than Standard German speakers (30 % as compared to 7 %, Ulbrich, 2004, 113). However, Ulbrich does not distinguish between nuclear and prenuclear occurrences of different accents types nor does she relate the accent types to function or meaning. Ulbrich's study provides only tentative evidence for the rise-fall appearing to be a characteristic of Southern German intonation. From a phonetic point of view, Ulbrich observes that the realisation of Swiss German rising accents starts from a higher pitch level to implement the accentual low tone as compared to Swiss German falling accents (cf. Figure 4.5). Note, that otherwise rising and falling accents are fairly similar in accent shape. Standard Swiss German appears to differ from Alemannic dialects in that it allows for nuclear falling accents (Ulbrich, 2004). Given that neither SwabG nor Freiburg Alemannic German allow for nuclear falling contours, this seems to be a significant difference between a Standard variety and a dialect. Turning to South-Eastern varieties of German, Barker (2002) provides a comprehensive account of the variety of Tyrolean German. Barker observes a nuclear rising pitch accent for Tyrolean German as it is the case for the South Western varieties of German. He claims that a rising nuclear pitch accent followed by a low phrase tone (L*+H L-) forms the default accent for Tyrolean German in a declarative sentence (Barker, 2002, chapter 6). However, Barker also analyses a simple falling contour that appears to compete with the rise-fall as being the neutral accentuation. In total, Barker claims four basic nuclear contours to exist in Tyrolean German, a fall (H*+L), 3

Note that Ulbrich (2004) gives no reference to accentuation in her notation of the rise-fall. Thus, the notation remains impressionistic.

53 a rise (L*+H), a fall-rise (H*+L H%), and a rise-fall (L*+H L-) (Barker, 2002, 195). Given that Barker's investigation is based on Fery (1993) and thereby on the theoretical assumptions of Gussenhoven (1984) (cf. chapter 3, section 3.2.5), the relation of these four basic contours to Gussenhoven's three basic intonational distinctions with respect to meaning remains undiscussed. Barker does not comment on that issue, a fact that appears to be unfortunate in the context of dialectal variation, in particular having in mind the intonational peculiarities of South-Western German varieties reviewed above. Barker (2002, 195) only notices that the fall and the rise-fall "have the implication of completeness or finality". Yet, from a systemic point of view, we do not know whether these contours are related, and thus variants of an underlying pattern, or whether these contours co-occur with certain differences in meaning. Table 4.1: A comparison of nuclear contours in Southern German that are analysed as neutral accentuation in the respective language variety.

Dialect South-Western German Our analysis Stuttgart SwabG Stuttgart SwabG Frey (1975) Fery (1993) SwabG Gibbon (1998) Cologne - Swiss German Freiburg Alemannic Gilles (2005) Freiburg Alemannic Peters (2006) Fitzpatrick-Cole (1999) Bern Swiss German Standard Swiss German Ulbrich (2004) South-Eastern German Barker (2002) Tyrolean German a b c d

Nuclear Contour L*H L% L*H L% a L*HL L*+H b L*+H HL- (%) L*+H Hfc+L (Lt) L*+H L% LH L(%) d

c

L*H L-

Our tonal adaptation, see (1) above. Note that Gibbon (1998) only refers to the nuclear accent. The diacritic L corresponds to %, see Gussenhoven (2004). Note that no reference to phonological accentuation is given.

In summary, while in Northern Standard German an intonational falling accent pattern functions both as default accent and as focus accent realisation (e.g. Fery, 1993), in Southern German inverted tonal relations are claimed to exist (Sievers, 1912). Table 4.1 summarizes the individual nuclear contours that correspond to a neutral intonation contour in the different dialects reviewed above. It is remarkable that researchers generally agree that Southern German intonation appears to exhibit a nuclear rising accent as default accentuation (cf. Table 4.1).4 As can be seen in Table 4.1, intonational differences between Southern German dialects seem to arise in the formation of nuclear contour type, i.e. the dialects differ in the realisation of the postnuclear pitch contour. SwabG - our object of study - appears to differ from Alemannic German with respect to the Alemannic postaccentual high plateau. As we will show in section 4

Note that Truckenbrodt (2004) also analyses falling accents (H+L*) in Southern German; see also Claßen (2002) for falling accents (H*L) in Swabian German.

54 4.2.2, in SwabG, the falling part of the rise-fall starts immediately after the pitch peak, which is more similar to Tyrolean German (cf. Barker, 2002). Yet, in contrast to the Tyrolean German low phrase accent considered as being part of the rise-fall, we propose a low boundary tone for SwabG since we observe that the fall aligns at the intonation phrase boundary independent of the position of the nuclear accent in the phrase.

4.2

Nuclear contours

In SwabG we distinguish between three basic nuclear contours, which differ from basic contours of the majority of other intonation languages or dialects. While for Standard German, English and Dutch as well as SaxG (chapter 5), the simple fall, the simple rise, and the fall-rise are assumed to represent the basic contours related to basic differences in meaning (e.g. Gussenhoven, 1984, 2004; Fery, 1993), in SwabG these are the rise-fall, the simple rise, and the rise-fall-rise, respectively. Additionally, we propose that a modified contour exists, which we colloquially refer to as the rise-falllow-plateau. These peculiar intonational characteristics of SwabG will be put into a general perspective of intonation, phonology and language in section 4.4. The following analysis of SwabG intonation patterns is mainly based on intonation phrases whose accented words contain the symbols of the map task since these contain sonorants to avoid microprosodic perturbations in the speech signal. We also show examples of accented words that contain less sonorant segmental material. Our analysis compares pitch accents in different positions in the phrase (early vs. late, cf. Grice, 1995). Based on Grabe (1998a), who found different surface accent realisations for phrase-final and non final accents of the same underlying tonal category, we also consider phrase-final and non-phrase-final pitch accents to account for pitch accommodation effects. We start our discussion of SwabG intonation with an overview of nuclear contours in section 4.2.1. Each of the nuclear contours is discussed separately, the rise-fall in section 4.2.2, the simple rise in section 4.2.3, the rise-fall-rise in section 4.2.4, and the rise-fall-low-plateau in section 4.2.5. Section 4.2.6 is concerned with falling accents, which as we argue do not belong to the core tonal grammar of SwabG. Section 4.2.7 discusses boundary specifications where we claim that SwabG differ from SaxG and Standard German, section 4.2.8 relates the rise-fall and the risefall-low-plateau contours to broad and narrow focus, respectively, and section 4.2.9, finally, summarizes the section on nuclear contours.

4.2.1

Overview of the data

This section intends to provide an overview of different nuclear contours produced in SwabG. From our map task corpus, we have chosen intonation phrases that contain the accented word Wohnwagen ['vo:n.,va.g8n] 'caravan' as the target. In total, the

55 corpus contains 23 instances of the target word in a nuclear contour. From these, eight are given in (3). In (3-a) and (3-h), the target word is non-phrase-final, while it is phrase-final in the other cases. The pitch tracks of the examples in (3) are displayed in Figure 4.6 showing only the nuclear syllable and its surrounding syllables. Obviously, different contours have been realised on the target word. Nevertheless, all contours shown in Figure 4.6 exhibit a low tonal target in the accented syllable followed by a sharp rise. Concerning the sharp rise our data is in line with the description of Stuttgart Swabian German of Frey (1975). (3)

Intonation phrases with a nuclear accent on Wohnwagen 'caravan'. Capitals indicate accentuation. a. b.

un na kommt bei mir der WOHNwagen jetzt da oben 'and then I now have the caravan up there' bis zum WOHNwagen 'until the caravan'

c. d. e. f. g. h.

ich hab noch η WOHNwagen Ί have yet a caravan' ach so da hasch du η WOHNwage 'oh dear, there you have a caravan' bei dir=isch=s ä lamm bei mir η WOHNwage 'for you it's a lamb for me a caravan' dann kommt bei mir η WOHNwagen 'then I have a caravan' da der WOHNwagen 'there the caravan' isch=d isch da boi dir=a WOHNwage da obe dranne 'do do you have a caravan up there'

The differences between the contours in Figure 4.6 concern the course of pitch in the postnuclear unaccented syllables until the end of the intonation phrase. Considering the first four pitch tracks (Figure 4.6a, b, c, and d), pitch falls after the end of the sharp rise that goes through the accented syllable, and the intonation phrase boundary is low. What distinguishes Figures 4.6a and b from Figures 4.6c and d, is the steepness of the fall, or, in other words, the alignment of a low tonal target with respect to the postnuclear string. In the former two cases (Figures 4.6a, b), pitch descends slightly, and the fall reaches its turning point at the end of the intonation phrase, i.e. interpolation of pitch occurs between the accentual high turning point and a low intonation phrase boundary. In the latter two cases (Figures 4.6c, d), the fall is steep and reaches its turning point in the first postnuclear syllable. After the low turning point, pitch levels out until the end of the intonation phrase. Figures 4.6e, f, and g show no fall after an accentual high turning point. Instead, pitch continues to rise in Figures 4.6e and f, and levels out forming a high plateau in Figure 4.6g. And finally, in Figure 4.6h, pitch falls rapidly as in Figures 4.6c and d, continues on a low level and rises again at the end of the intonation phrase.

56

\



i

i i

1

i

! Ii —Γ

i i Ii I I

h ie WOHN

200

i i i ! I i i i

1 \1

1 ! i 1 I

wa ge jetzt da

jj i /

-

\

h ^

obe

zum

WOHN

i \ | i j 1 i wa

| | | 1 1| ^ j 1 i

(b)

gen

(d)

500

300(f)

150WOHN

wa

en

ge

WOHN

wa

gen

400 250(h)

150-

Figure 4.6: Comparison of different pitch contours on the target word Wohnwagen 'caravan' and its surrounding syllables. The numbering of the pitch tracks corresponds to the numbering in (3).

The preliminary phonetic analysis of Figure 4.6 suggests a five-way distinction of nuclear pitch contours. We can distinguish a rise-fall from a rise-fall-low-plateau (Figures 4.6a, b vs. 4.6c, d), a simple rise from a simple rise-plateau (Figures 4.6e, f vs. 4.6g), and a rise-fall-rise (Figure 4.6h). The overview of Figure 4.6 does not show any instance of a fall or fall-rise. These accents are apparently occurring in Standard German (e.g. Fery, 1993), yet we will argue throughout the following sections that falling accents do not belong to the core tonal grammar of SwabG. Other contours of Standard German appear to occur in SwabG as well, cf. the rise-fall and the simple rise (Fery, 1993, 93ff and 85ff, respectively for Standard German).

57 In the following, we claim that SwabG intonation consists of three basic intonation contours, i.e. the rise-fall (Figure 4.6a, b), the simple rise (Figure 4.6e, f, g), and the rise-fall-rise (Figure 4.6h). This means that the simple rise and the simple riseplateau contours are treated together, since a difference in function between the two is not apparent for us (cf. section 4.2.3). Concerning the rise-fall (Figure 4.6a, b) and rise-fall-low-plateau contours (Figure 4.6c, d), we analyse the latter as being a modification of the former (in the sense of Gussenhoven, 1984). Given their structural similarity - the crucial difference is in the alignment of the postaccentual low tone - we observe that the two contours differ in function: according to our analysis in section 4.2.8, the rise-fall-low-plateau contour tonally marks narrow focus, whereas the rise-fall marks broad focus.

4.2.2

The rise-fall

In this section, we will argue that the rise-fall is best accounted for by analysing it as a sequence of a rising pitch accent plus a low boundary tone (L*H L%, cf. Figure 4.7). We will show that the shape of the fall depends on the position of the rising accent in the phrase, that is, the fall always aligns at the end of the intonation phrase irrespective of the distance between the rising pitch accent and the intonation phrase boundary (Figure 4.7). Further, we will show that the rising accent shape is identical when comparing its realisation in different positions of the phrase. The rise-fall is by far the most common tonal pattern in SwabG (cf. Table 2.2, chapter 2). It is realised in statements, such as declaratives, requests, complex sentences with subordinate clauses and the like, and in yes-no-questions as well as wh-questions.

Figure 4.7: Nuclear rising pitch accent with a low boundary tone schematically represented on two, three and η phrase-final syllables; phrase-finality is indicated by a square bracket; each box represents a syllable, shading refers to accentuation.

Phonetically, the rise-fall contour is realised as follows. The accentual low tonal target occurs very early in the accented syllable (on average 2.9 ms prior to the syllable nucleus, cf. chapter 6, section 6.2.3). It is characterized by a slight dip in pitch from a preceding higher pitch level (cf. Figure 4.7). This realisation corresponds to what Ulbrich (2004) found for Standard Swiss German rising accents (see Figure 4.5 above). From the low target, pitch rises sharply to a high target that is reached near the end of the accented syllable (on average 68.5 ms prior to the end of the syllable rime, cf. chapter 6, section 6.2.3). 5 After the rising pitch accent, pitch falls gradually until the end of the intonation phrase, which is indicated by a low boundary tone. 5

The high target may also be realised after the accented syllable (Kügler, 2004b).

58 The phonological structure of the rise-fall is L*H L% (cf. Figure 4.7). We analyse the rising pitch accent with an accentual low tone L* and not with an accentual high tone (H*), since the low tone appears perceptually more prominent than the high tone. The high tone, we analyse as a trailing tone, since it aligns fairly stable in the accented syllable. If we were to analyse it as a high phrase accent, we would not expect that the realisation of the high tone is governed by the accentual low tone. Instead, we would expect a rather variable alignment pattern dependent on other factors like, for instance, the occurrence of a postnuclear stressed syllable or the distance to the end of the intonation phrase (cf. Grice et al., 2000, for Η phrase tones in Standard Romanian and Standard Greek, respectively), which we do not observe in our data. As Figure 4.7 suggests, the fall depends on the intonation phrase boundary, which lead us to analyse it as low boundary tone. According to Figure 4.7, the shape of the rise-fall contour depends on the number of syllables that follow the accented syllable. We will examine this in detail now by looking at rise-fall contours where the distance of the nuclear accent to the end of the intonation phrase varies. In phrase final words containing at least two syllables with word stress on the penultimate syllable, the rise-fall is realised on the accented and postaccented syllable (cf. Figure 4.8a and b). Consider the examples of Figure 4.8a and b in (4) where the accentual low is associated with the stressed syllables SEE and MEHL. From the low turning point, the pitch rises sharply to the high trailing tone. According to Pierrehumbert (1980), a tone is associated with a syllable, and each tone receives its own docking site if enough syllables are available. Therefore, we assume that the trailing high tone is associated with the following unstressed syllable, as shown in (4). Recall, however, that the actual alignment of the trailing tone is on average on the accented syllable. On the phrase-final syllable, the pitch falls down to the bottom of the speakers' range which is accounted for by a low boundary tone that associates with the right edge of the intonation phrase, i.e. the last syllable in the phrase. In (4), therefore, the postaccentual unstressed syllable has two tonal associations. The examples in (4) are like the one described by Frey (1975) and Fery (1993) for SwabG, see section 4.1. (4)

a.

ein äh SEEle

'a uh small lake'

b.

L* Η L% isch en MEHLsack

'is a sack of flour'



I /I

L*H L% If the nuclear accent is realised earlier than on the penultimate syllable of the phrase, we observe that the fall in pitch is less steep (cf. Figure 4.8c and d with their tonal association in (5)). The shape of the rising accent is identical to the one before (Figure 4.8a and b). The falling part of the rise-fall extends to the postnuclear syllables. In (5-a), each of the three tones receives its own syllable to associate with. The accentual low associates with the stressed syllable of the nuclear words WOHNwagen 'caravan' and NONnenweiher 'idyllic pond'. The trailing tone associates with the first postnuclear syllable, and the boundary tone with the phrase-final one.

59

Ο

0.2

0.4

0.6

0.8

1

Ο

0.2

0.4

s, —

0.6

0.8

· — ·\

,.·—• —.

\ I L* I bis Ο

WOHN

zum 0.2

I Η 1

0.4

0.6

I L% 1 wa

beim

gen

0.8

1

1

1 L* 1

Ο

ΝΟΝ 0.2

1

Η 1

L% 1

weiher

nen 0.4

0.6

0.8 0.9

Λ

Λ

^^

L* 1 Um Ο

ι Η ι den 0.2

ι L* ι

ι Η ι

MEHL 0.4

0.6

ι L% ι sack 0.8

ι Η* ι

rum

1

1.1

un=na Ο

ι L* ι

0.8

•Ν ν. I L% ι

ι !Η ι

I omrr bei mir d e l W O H N 0.4

%

wa

1.2

ge

jetzt d a o b e

1.6

2

2.3

Figure 4.8: SwabG rise-falls with the accent realised in different positions in the phrase but not phrase-final.

(5)

a. b.

bis zum WOHNwagen L* Η L% beim NONenweiher L* Η L%

'until the caravan' 'at the idyllic pond'

If the rise-fall is associated with a non-phrase-final word, the shape of the accent is again identical to accents before (cf. Figure 4.8e and f). The shape of the contour differs, though, since the gradual fall is slower than before. Interpolation of pitch occurs between the high trailing tone and the low intonation phrase boundary, cf. the tonal association in (6), where the number of postaccentual syllables ranges from two to six. The longer the postaccentual stretch, the smoother the postaccentual fall.

60 In (6-b), the fall extends to six postaccentual syllables, and pitch falls slowly, which might lead to the impression of a plateau-like postnuclear stretch in Figure 4.8f. (6)

a.

Um den MEHLsack rum 'around the sack of flour' L*H L* Η L% un na kOmmt bei mir der WOHNwage jetzt da obe H* L*H L* !H L% 'and then comes the caravan up there on my (map)'

b.

With these examples, we have presented evidence that the falling part of the risefall contour is independent of the accentual rise in that it aligns at the intonation phrase boundary irrespective of the occurrence of the nuclear pitch accent. We have also observed that the shape of the pitch accent is identical in different positions of the phrase. The next part of this section examines rise-fall contours in phrase-final positions. If the rise-fall occurs on phrase-final syllables, the pitch accent as well as the boundary tone associate with the same syllable as in (7). The implementation of three tonal targets causes the pitch accent to shift its high trailing tone leftwards, i.e. the high tone is realised earlier in the accented syllable in order to leave space to realise the final fall (cf. Figure 4.9).

\

.

1 L* 1

0.2

1 L* 1

1 1 H*L% 1 1

des=sch bloß ä 0

r

A

LAMM

0.4

0.6

\ Ν•χ

0.8 0 9

Λ A

100 I L*H

ι ι L* !Η I

ι

ι

0.2

0.4

0

200

\

bei len schrift

\ I L%

130

I

0.8

1

0.3

nOn 0.6

0

nen

0.9

I I L*H ΛΙ_°/
·

ν



V A

\ \ L*

L*!H I

L%

1 L*> 1

I Η I

en MEHL

sack

sEn

krecht

Η

unlda

stAht

0.3

0.6

0.9

1.2

1.5 1.7

: :

I

I L* !H I I

II L*!H II

1 ~L% 1

hOch zum WOHN wa 0.8

0.4

>

1.2

ge

1.6

Ν

/

/ ι

ι

ι

ι

I

L '

Η

~ L %

L*

Η

1

1 R U N

0

0.1

1

1 ter

0.2

0.3

0.4

[da isch

0.5

(.)

0.4

/ 1

L

*

Η

I

rEchts

0.8

1 d a v o n

s o = n

1.2

V1EHL

1.6

1 Λ

Ι_%

1 s a c k

1.9

Figure 4.10: Phonetic variants of rise-falls; upper panel: downstep of the trailing tone; lower left panel: phrase final creaky voice; lower right panel: non-completed final fall.

62 (8)

a. b. c.

un dA staht en MEHLsack 'and there is a sack of flour' L*H L*!H L% sEnkrecht hOch zum WOHNwagen 'vertically up to the caravan' L*H L*H L*!H ~L% rEchts (-) RUNter 'right downward' L*H L*H ~L% da isch (.) rEchts davon so=n MEHLsack 'on the right of it, there L*H L*H ~L% is like a sack of flour'

The phonetic adjustment "truncation" of tones is well known in some languages. In Hungarian, for instance, a question is associated with the tonal sequence of L* Η L%, where the Η tone consistently associates with the penultimate syllable of the phrase and the L% with the final one (Grice et al., 2000; Ladd, 1996, 115-118). If the question tune has to be realised on a phrase-final syllable, e.g. on a monosyllabic one-word phrase, the final fall is not realised at all. On the surface, only a simple rise remains (Ladd, 1996, 132-133). Although the Hungarian case differs from SwabG where the falling part is initiated, the same phonetic mechanism applies. The second phonetic variant concerns phrase-final syllables that have been realised with creaky voice (about 8 % of the cases), i.e. no pitch trace is available for analysis. However, creaky voice is an indicator for low pitch, the vocal cords are vibrating very slowly. In accordance with the International Phonetic Alphabet (IPA, 1999), we label a tone realised with creaky voice with a tilde as a diacritic (cf. (8-b) and (8-c)). The third surface variant of a rise-fall concerns the phonetic implementation of the final fall. In (8-d), the nuclear rising pitch accent is realised on the accented syllable. The final fall, i.e. the expression of the low boundary tone, is truncated. This means that the fall does not end at the bottom of the speakers' range but remains on a slightly higher level. Acoustically, the falling gesture is initiated which causes the perceptual impression of a fall. Thus, the truncated final fall is a phonetic variant of a low boundary tone. An intonation phrase containing a truncated final fall exhibits falling intonation globally and may not be confused with a phrase that ends in a pitch plateau (cf. section 4.2.3). The higher pitch level of the low boundary tone is labelled with a " as a diacritic (cf. Grice and Baumann, 2002, for this labelling convention). Yet, phonologically this tone is a low boundary tone (L%). If the final low is truncated, the result is always a case of tone clash, i.e. when the accent and boundary tone clash on a phrase-final syllable. In very few instances, a truncated fall may be observed on phrase-final trochees. We observe an interesting correlation between downstep and truncation of the final fall: if the nuclear accent is downstepped on a phrase-final syllable (cf. Figure 4.9c, d), no truncation of the final occurs. A possible explanation for this could be that the lower scaling of the downstepped pitch accent allows for a realisation of a complete fall since the fall simply starts from a lower level. However, the data set is too small to generalize a correlation between accent type and realisation of the fall. This should be verified in an experimental setting. The examples presented so far resemble declarative sentences. If we consider questions, we observe the same tonal pattern for yes-no-questions (cf. Figure 4.11, left panel). In a wh-question - the only one with a rise-fall in our corpus (cf. Figure 4.11,

63 right panel) - we observe a tonal pattern similar to the Alemannic type described by Gilles (2005) and Peters (2006). In Alemannic, a rise-fall is typically realised with a high plateau after the rising pitch accent and prior to the final fall (see section 4.1 above). Since SwabG and Alemannic German are closely related, the similarity in tonal structure is not unexpected. However, we cannot verify whether this pattern is a tonal characteristic of wh-questions in SwabG. Therefore, further investigation is needed to determine whether or not wh-questions are realised with a rise-fall in SwabG. (a)

200

Χ

300

100 70

(b)

450

V

200

L* Η 1 1 isch des bei dir AU heller 0.3

0.6

0.9

150

L% 1 diamant 1.2

ι ι L* Η ι ι WIE heißt des bei 0.2

1.51.6

0.4

0.6

ι L% ι dir 0.8 0.9

Figure 4.11: SwabG rise-fall on questions; (a) yes-no-question, (b) wh-question. The tonal association of Figure 4.11a is given in (9). As for the examples above, the accentual low is associated with the accented syllable, the trailing tone with the postaccentual and the boundary tone with the final one. The fall extends to five unaccented postnuclear syllables. (9)

isch des bei dir A U heller diamant L*H

'do you also have bright diamond'

L%

Concerning the function of the rise-fall, we will see in section 4.2.8, that the rise-fall expresses broad focus tonally. Given the frequency and the distribution of the risefall, we assume that this contour is the default accentuation in SwabG, which implies that we analyse the rise-fall as a base contour. Given that, we have to distinguish it from the rise-fall of Standard German described by Fery (1993). As we have argued above (section 4.1), the Standard German rise-fall should be interpreted as a modified version of a base contour. Given Fery's analysis, the rise-fall appears not to be a base contour of Standard German. Even if it were, the distribution of the rise-fall is very restricted (Fery, 1993) and not equivalent to that of the rise-fall in SwabG. We therefore believe that the SwabG rise-fall as a base contour has a different phonological status in the tonal grammar than the Standard German rise-fall. The two are similar in form, yet the functional differences are evident. Summary The rise-fall contour is the most common contour in SwabG occurring in declaratives as well as in questions. It is composed of a nuclear rising accent and a low boundary tone. We have argued that the postaccentual fall does not be-

64 long to the accent since its shape depends on the position of the accent in the phrase, that is on the number of following postnuclear syllables. Interpolation of pitch occurs between the accent and the phrase boundary. The rising accent is characterized by a slight dip in pitch towards a low tonal target at the beginning of the accented syllable. The low tone is followed by a sharp rise in pitch towards a high trailing tone. The phonological form of the rise-fall in terms of the tone-sequence model is L*H L%. The intonational function expressed by the rise-fall is broad focus.

4.2.3

The simple rise

In this section, we will examine the second base contour of SwabG, the simple rise, which is almost as common and frequent as the rise-fall discussed in the previous section (cf. Table 2.2 in chapter 2). Figure 4.12 displays the simple rise contour schematically.

• • C Z D I Z Z ] I I I L* Η H% Figure 4.12: A simple rise in SwabG schematically represented in a three syllable environment; the dotted line represents a variant of the simple rise; the grey shaded area represents word stress.

A simple rise occurs frequently in declaratives as well as in questions. In either case, the simple rise consists of a rising pitch accent of the form L*H. After the nuclear accent, pitch levels out until the end of the intonation phrase on the pitch level of the high trailing tone (dotted line in Figure 4.12), or rises on the final syllable of the intonation phrase (solid line in Figure 4.12, labelled as H%). These two variants of the simple rise are phonetic options, since we do not find any apparent difference in intonational meaning. We thus comprise the contours L*H and L*H H% under the simple rise contour. 6 In the following presentation of the data, we will show the distribution of the simple rise contour in different structural contexts, i.e. with the nuclear accent in different positions in the phrase. We start with declarative sentences first, and turn to the simple rise in questions below. We first consider different accent positions within the phrase in order to show that the contour shape of the simple rise differs as a function of the position of the nuclear accent in the phrase. Figure 4.13 below gives different examples of simple rises, with their autosegmental association given in (10). In Figure 4.13a, b, simple 6

Note that Peters (2006) analyses these two rising contours in Freiburg Alemannic German as two distinct base contours. Peters refers to the L*H H% as the double rising contour, and to the L*H contour as the rising-plateau contour.

65 rises whose rising pitch accent is associated with a phrase-final trochee are shown. T h e pitch rises through the accented syllables of the words LAO jet 'walking' and FLUgel 'wing' ((10-a) and (10-b)). In Figure 4.13c, d, the rising accent associates with the antepenultima (or earlier) syllable of phrase-final word. T h e rise is realised on the words WOHNwagen 'caravan' and NONnenweiher 'idyllic pond' ((10-c) and (10-d)). A n d in Figure 4.13e, f, the rising accent associates with a non phrase-final word ( M E H L 'flour' and AU 'also', (10-e) and (10-f)).

(a) V

χ/

200 150

/

200

0.4

I L*H j

fet

0.6

0.8

1

1.2

)

0.2

I L* ι

lEtsch te

bis zum

(c)

y

y

V

1 Η 1

LAU

0.2

/

300

1 L* 1 und

(b)

500

0.4

0.6

FLÜ

gel

0.8

1

1.2

-

(c)

wagen

zum

ι i

200

i i

150

i

i

-l -

(b)

i

WOHN

wagen

hin

(d)

ri ^ v A

100-1

100 jetz=n

WOHN

wagen

150

(e)

WOHN

150-

/

I-

100-

/"Y

wagen

\ (f)

10070 mir 250-

WOHN

wagen

250

200-

200

(g)

(h)

100

100 WOHN

wagen

WOHN

wagen

Figure 5.5: Comparison of different pitch contours on the target word Wohnwagen 'caravan' and its surrounding syllables. The numbering of the pitch tracks corresponds to the numbering in (1).

the overview of nuclear contours in Figure 5.5, since we did not find this contour on the word Wohnwagen 'caravan'. The fall-rise occurs rarely in our corpus. Following Gussenhoven (1984), we distinguish between base contours and modified contours. Gussenhoven's claim is that a tonal modification applies to all base contours and results in similar tonal surface realisations. Our SaxG analysis proposes that the simple fall (5.5a, b, c), the simple rise (5.5g, h) and the fall-rise are analysed as the three basic nuclear contours of SaxG. According to Gussenhoven (2004) the

102

modification is claimed to be L-prefixation, which affects the simple fall and the fallrise resulting in a rise-fall (5.5e) and a rise-fall-rise (5.5d, f), respectively. L-prefixation causes an additional low tonal target prior the nuclear accent, which is identical for all nuclear base contours with the exception of the simple rise; a modified simple rise did not occur in our data, yet we would expect it to exist in SaxG intonation.

5.2.2

The simple fall

In this section, we will argue that the simple fall is best accounted for by analysing it as a falling pitch accent without any further boundary specification (H*L, cf. Figure 5.6). We will show that the tonal shape of the accent does not depend on the position of accent in the phrase, that is, both accent tones are realised in relation to each other. The simple falling accent seems to be the most frequent and natural pattern in SaxG intonation for declarative sentences (cf. Table 2.2, chapter 2) in contrast to SwabG but equivalent to Standard German (Fery, 1993). Generally, the simple fall conveys a meaning of finality and completeness, and this is also true for SaxG, yet the simple fall may occur in yes-no-questions and wh-questions as well, which we will show below.

CZZ3CZIC I H* L Figure 5.6: Schematical representation of a SaxG nuclear simple fall; shading indicates word stress. Phonetically, the high tone of the accent occurs early in the accented syllable (on average 15.7 ms after the beginning of the syllable nucleus, i.e. on the vowel, cf. Table 6.2 in chapter 6). The fall extends to the accented syllable and reaches its low turning point on average 65.9 ms after the accented syllable (cf. Table 6.2 in chapter 6). We also identify many instances where the low tone is realised in the accented syllable, more precisely at the end of the accented syllable. This variation of the realisation of the low tone can be seen in the relatively high standard deviation of 141.6 in Table 6.2 in chapter 6. After the low tone, pitch levels on the remaining syllables, if any. The phonological structure of the simple fall is H*L. We analyse the high tone as the most prominent tone of the accent since it is perceptually more salient and acoustically more invariant than the low tone. The low tone is analysed as a trailing tone. Our data does not suggest a stress-seeking behavior of the low tone that would indicate that the low tone should be analysed as a phrase accent in the sense of Grice et al. (2000). The low tone is often realised in the accented syllable itself which would not be expected of a phrase tone. Since the pitch contour stays rather low after the accentual fall (if enough syllables follow the accented one), we assume that the intonation phrase boundary is tonally unspecified (cf. Figure 5.7). Following Fery

103

220

200

150

" "" κ (a)

100

\

I Η* 1

0.3

0.6

0.9

0.2

L ι EN

d a steht der

1.5

1.2

ι

ι Η* ι

1 ten

\

04

gel 0.6

Ν \ /V

(c)

300

_ -Ν

100

1 L

UN

n a d e r liegt r e s c h t s

(b)

300

(d)

200

Λ

150

1

I

L'H 1 unc

H* 1

dA

s t e h t ein

0.3

0.6

EN 1.2

(e)

300·

a« I L ι I

ι H*

L* ι un=machst dann=n

gel

0.9

\

\ -

ι I Η I

100

I L 1

0.7

1.5

0.3

krEuz 0.6

ι nd schreibe Z I E L 0.9

1.2

daneben 1.5

1.7

(f)

220-

200

200

150

150



100

ι H* ι SUMMT

die 0

100

I L I

0.1

0.2

0.3

noch 0.4

0.5

SAN

dieses 0

0.3

ge 0.6

rin

ϊγ

se 0.9

1.2

1.5

1.8

Figure 5.7: SaxG simple falls with the nuclear accent realised in different positions in the phrase but not phrase-final.

(1993) and Grabe (1998a), we account for a low postaccentual pitch plateau assuming that the low tone spreads until the intonation phrase boundary (cf. Figure 5.6). T h e phonological structure of the simple fall is identical to the Standard German simple fall (Fery, 1993; Grabe, 1998a), yet it differs phonetically. The simple fall is illustrated in Figure 5.7 with realisations on non-phrase-final syllables. Examples of a simple fall on phrase-final trochees are given in Figures 5.7a, b and c. In the first two examples, the intonation phrases contain only one pitch accent, and the fall starts early in the accented syllables of the phrase-final words UNten 'below' and ENgel 'angel'. T h e phrase in Figure 5.7c contains a prenuclear rising accent. In Figures 5.7a and c, the trailing low tone is realised in the postaccentual syllable, whereas it is realised in the accented syllable in Figure 5.7b. T h e tonal association, given in (2), follows the principles of association according to Pierrehumbert (1980),

104 where each tone receives its own docking syllable if there is any. Thus, in (2), the accentual high tone is associated with the accented syllables, and the trailing low tone with the postaccentual ones. (2)

a.

na der liegt reschts UNten H*L~

b.

da steht der ENgel

c.

H*L und dA steht ein ENgel

I/

L*H

'well that is below on the right'

'there is an the angel'

'and there is an angel'

I/

H*L

If the nuclear falling accent occurs earlier in a phrase, it exhibits the same shape and association as in phrase-final trochees, illustrated in Figures 5.7d, e and f. The number of postaccentual syllables ranges from one (Figure 5.7e) to three (Figure 5.7d). Since these examples show no difference in accent shape, we analyse the tonal association in (3) as above, yet tonal spreading of the low tone is shown. That is, the low phrase boundary is accounted for by means of low tone spreading. (3)

a.

un=machst dann=n krEUz und schreibst ZIEL daneben

b.

L*H H*L 'and then make a cross and write finish next to it' die SUMMT noch 'it is still humming'

c.

H*L schließt dann (-) dieses SÄNgerinnenseminar a ein

I/

I

H*L 'includes the seminar of female singers' With the examples so far, we have presented evidence that the accent shape does not differ if realised earlier in an intonation phrase. In particular, the low trailing tone appears to be realised in relation to its accentual high tone. Its realisation neither relies on positional effects nor on the occurrence of postnuclear stressed syllables (cf. Figure 5.7), factors that might be evidence for assuming a phrase accent (Grice et al., 2000). Thus, we analyse the low tone as a trailing tone. If we consider nuclear simple falls on phrase-final syllables, we observe that the accent shape is similar to the previously presented ones. The fall starts early in the accented syllable and ends at the end of the intonation phrase, thus implementing two tones on one syllable. Examples are given in Figure 5.8. The tonal association is given in (4). Both tones associate to the accented syllables, respectively, since no further syllables are available for the tones to dock onto.

105 (4)

a.

sUch isch mir mal s beste R A U S

b.

H* H*L gut also wOhnwagen heißt jetzt L A M M

c.

L*H und dann bin ich am ZIEL

I

f\

\

'and then I am at the finish'

H*L

w O liegt der sEe ungeFAHR L*H

H*L

'where is the lake approximately'

!H*L

(a)

200·

(b)

160·

V

150

\ 1/

100 100-

1 Η* 1

1 H* 1 isch mir mal s beste

such 0.2

0.4

'ok, caravan is called lamb now'

!H*L

Κ

%H d.

Ί choose the best one'

0.6

70·

1 L 1

gut also

RAUS 0.8

1

1.1

0.3

/

I

1

L*

Η

1

1

WOHN 0.6

ι !Η* ι wa

heißt jetzt

|ΘΙ

0.9

1.2

ι L ι

LAMM 1.5

Figure 5.8: Simple falling nuclear accent on phrase-final syllable. Target words are RAUS 'out' (a), LAMM 'lamb' (b), ZIEL 'finish' (c), ungeFÄHR 'approximately' (d). If we consider the tonal implementation of simple falls on phrase-final syllables, it is constrained by the upcoming intonation phrase boundary. The two tones are realised on the accented syllable. Since this is also true for non-phrase-final accents, we conclude that in SaxG simple falls, the position of the accent in the phrase does not affect the shape of the contour. This is in contrast to what Goldbeck and Sendlmeier (1988) and Grabe (1998a) found in Standard German: The accent shape in phrasefinal position differs from that in non phrase-final position.

106 If compared to Standard German, the SaxG accent shape of the simple fall differs from Standard German with respect to tonal alignment. Grabe (1998a) showed that the nuclear fall in Standard German aligns the pitch peak at the right edge of the syllable rhyme. Our SaxG data show that peak is consistently aligned at the beginning of the stressed syllable's rhyme, i.e. near the beginning of the vowel (cf. chapter 6). So SaxG aligns the beginning of the fall earlier than Standard German. Gilles (2005) showed that the alignment of the peak in falling accents differs in German varieties. Thus, the result here is further evidence of dialectal variation with respect to tonal alignment. The simple fall is frequently used in hat patterns where it is the second accent of two accents in a sequence. This is in line with Standard German (Fery 1993, and Wunderlich's 1988 accent C "Fallend-Tiefakzent", cf. Figure 3.4 in chapter 3). An example has been shown in Figure 5.7c above. Further examples of the hat pattern are given in Figure 5.9. In Figure 5.9a and b, the simple fall is associated with a phrase-final trochee, whereas in Figure 5.9c with a non-phrase-final word. As for the above presented examples, the shape of the contour does not differ for accents in different positions. The autosegmental association of the hat patterns is given in (5). (5)

a.

und da kOmmste bei der irAnischen arMEE (-) vorbei

I/

L*H

n

IH*L

L*H

'and then you are passing t h e Iranian army'

b.

Über die libelle DRÜber

c.

L*H !H*L 'above the dragon fly' dann erscheint plötzlich eine sUmmende liBELle

I/

I/

L*H !H*L 'then suddenly a humming dragon fly appears'

So far, we have presented intonation phrases that resemble declarative sentences. Example (4-d) above, however, shows that this accent pattern also occurs on whquestions. The usage of the simple fall for both declaratives and wh-questions has also been found for Standard German (Fery, 1993, 85). Another example of a whquestion with a simple fall on a non-phrase-final syllable is given in Figure 5.10. Again, the fall starts early in the accented syllable DIE 'they' and reaches its end at the end of the accented syllable. The course of the fall is interrupted at the right edge of the word and pitch continues falling slightly on the phrase-final word. See (6) for tonal association. To conclude, the occurrence of the simple fall as a nuclear pitch accent is not restricted by sentence type. (6)

was wOllen denn DIE hier \ y H*L

/ / !H*L

'what do they want here'

107

(b)

(C)

150

\

100-

\

\

1 1 L* Η I I Ü ber 0.2

die libelle 0.4

0.6

I L I ber

1 !H* I DRÜ 0.8

1

80

·. L* h 1 sUm men de

1.1

0.2

Ii 0.4

Ή' ι BEL 0.6

L ι Ιβ 0.8

0.9

Figure 5.9: Simple falling nuclear accents in hat patterns. Target words are arMEE 'army' (a), DRÜber 'above' (b), and UBELle 'dragon fly' (c).

200/

150-

100-

0

ι ι ι Η* L Η* ι ι ι wOl l n denn DIE was 0.2

0.4

0.6

hier 0.8 0.9

Figure 5.10: Simple falling nuclear accent wh-question. We observe two different surface variants of the canonical simple fall contour: downstep of the accentual high tone (Figures 5.8b and d, and 5.9b and c) and creaky voice on intonation phrase-final syllables (Figure 5.11, left panel). Creaky voice on phrase-final syllables is indicated by a tilde preceding the tonal label carrying over the IPA conventions, which is illustrated in (7). Creaky voice usually indicates low pitch, thus a variant of a low tone.

108

200

\ 1 H* I

ν—

\ 1 !H* 1

bis

zum 0.2

\

100·

\ 1 L 1

50

WOHN 0.6

0.4

gen

wa

1 H*L I

I !H*L I

über

die

0.8 0.9

0.5

sUmmende

I !H*L I libelle

I !H*L I Η

1.5

hin 2.5

weg 2.8

Figure 5.11: Phonetic variants of simple falling nuclear accents: Creaky voice (left panel), and total downstep (right panel). (7)

bis zum WOHNwagen

I

I

Η*

H*

'until the caravan'

1/

~L

Downstep affects the accentual high tone in relation to a preceding high tone, and is indicated by an exclamation mark, as illustrated in (5-b) and (5-c) above. With respect to downstep, Grabe (1998a, 89ff) showed for Standard German that accents may be partially or completely downstepped (cf. Figure 5.12). There is no categorical - only gradient - distinction between the two. In our SaxG corpus, we observe the same variation. We have presented partially downstepped accents above (e.g. Figures 5.8b and d). An example of a completely downstepped accent is shown in Figure 5.11, right panel, with the accent on hinWEG 'away'. The accent in Figure 5.11 might be analysed as a low tone, since the pitch is just low on the accented syllable.12 However, despite the difference in intonational form, we observe no difference in intonational function between a completely downstepped accent as in Figure 5.11 or a partially downstepped accent as for instance in Figure 5.8b, d above. Thus, we assume, in line with Grabe (1998a), that downstepped accents offer gradient variants with respect to the amount of downstep.

(a) —

I

!H*

I L

(b) I !H*

I L

Figure 5.12: (a) partial and (b) total downstep according to Grabe (1998a, 89).

12

Note that GToBI provides a distinction between H * + L and H + L * accents: The latter one seems to reflect the low pitched completely downstepped accent. However, Grabe (1998a, chapter 6) successfully showed that no categorical distinction can be drawn between ! H * + L and H + L * .

109

Until now we have shown that the shape of the simple fall is realised consistently in different positions of the phrase, and that this contour is used in declaratives as well as in wh-questions. With regards to the linguistic function, in section 4.2.8 below we argue that broad focus is expressed by means of the simple fall. We further assume that the simple fall conveys a meaning of 'confirmation of established mutually shared knowledge between speaker and hearer' (cf. Gussenhoven, 1984; Pierrehumbert and Hirschberg, 1990). In chapter 7, we apply Gussenhoven's model of intonational meaning and attempt to show that the simple fall conveys the meaning of A D D I T I O N (Gussenhoven, 1984).

Summary The simple fall contour is the most frequent thus most common intonation pattern in SaxG occurring in declaratives as well as wh-questions. The simple fall is composed of a nuclear falling pitch accent, which is schematically shown in Figure 5.6 above. The shape of the accent is independent of its position in the intonation phrase. The accentual high tone of the simple fall aligns with the left edge of the accented syllable's rhyme; the realisation of the low tone varies, in 57 % of the cases the low tone is realised in the postaccentual syllable. With respect to accent realisation, we have observed two surface variants of the simple fall, creaky voice at intonation phrase-final syllables and downstep. With respect to downstep, partial and complete downstep occurs. In the hat pattern, the simple fall is the second accent of the accent sequence. The phonological form in terms of the tone-sequence model is H*L with a tonally unspecified intonation phrase boundary. The trailing low tone spreads until the end of the phrase. We have argued that the accentual high tone is perceptually more prominent and shows acoustically a more invariant pattern than the low tone, therefore we analyse the low tone as a trailing tone. The intonational function expressed by the simple fall is broad focus.

5.2.3

The simple rise

In this section, we will present two types of simple rises. One is composed of a nuclear rising pitch accent plus a high boundary tone (L*H H%), the second consits only of a nuclear rising pitch accent without further boundary specification; in the latter case, we assume that the high tonal target spreads until the intonation phrase boundary forming a high plateau (L*H). The simple rise generally signals that a speaker intends to continue. According to our discussion of different sentence types in chapter 2, section 2.3.3.3, we assume a distinction of two different kinds of continuation. We will present the data separately for each type of continuation. From the distribution of the simple rise, it is not apparent to us that the two different contours differ with respect to function or meaning, thus we treat them as phonetic variants of the simple rise, which are shown schematically in Figure 5.13. Our data shows that the accent shape by and large does not depend on the position of the accent in the phrase. Figure 5.13 schematically presents two types of the simple rise, the variant with a high boundary tone (solid line) and without a boundary specification (dotted line). The simple rise occurs frequently in declaratives in our speech corpus, and is also used in questions. It is realised both in wh-questions, in yes-no-questions as well as in

110 declarative questions. The simple rise in questions is discussed below. We analyse the simple rise as L*H (H%). The alignment of the accentual low tone is fairly invariant, it aligns at the beginning of the accented syllables' rhyme (0.6 to 3 ms into the vowel, cf. Table 6.2 in chapter 6). The realisation of the high tone varies, in 50 % of the cases, it is realised in the accented syllable, in the other 50 % it is realised on the first postnuclear syllable. The invariant pattern of the low tone in combination with the perceptually salient impression of a low accented syllable leads us to assume an accentual low tone (L*), whereas the the high tone is analysed as trailing tone.

L_JI I I L* Η

IIZZI I H%

Figure 5.13: Schematical representation of a SaxG simple rise, the dotted line represents a variant of the simple rise; shading indicates word stress.

The data discussion here follows the distinction of two types of continuation made in chapter 2, section 2.3.3.3. We start with type I, list continuation.13 For list continuation, we identify both types of simple rises as Figure 5.14 illustrates. In all examples, the simple rise is realised on non-phrase-final syllables. In Figures 5.14a and b, instances of phrase-final trochees are shown, where pitch rises through the accented syllables RUNter 'down' and WINkel 'angle'. In Figures 5.14c and d, two cases of rises are shown, where the accented syllable is part of the phrase-final word, but (main) word stress is prior to the penultimate syllable (TRAUerweide 'weeping willow' vs. WOHNwagen 'caravan'). Finally, Figures 5.14e and f display two examples, where the rise is realised on a non-phrase-final word (diaMANT 'diamond' vs. ENgel 'angel'). The autosegmental association is given for Figures 5.14c and d, in order to illustrate the tonal difference in the postnuclear part of this contour. In (8-a), Figure 5.14c, the association of a high boundary tone with the final syllable of the phrase is shown. In (8-b), Figure 5.14d, on the other hand, spreading of the trailing tone is shown with the result of an unspecified phrase boundary. (8)

a.

b.

13

da steht ne eh ne TRAUerweide I L*H H% is bei mir direkt der WOHNwagen I L*H

I

\

'there is a weeping willow'

Ί have directly the caravan'

Recall that lists refer to a conversational activity, cf. chapter 2, section 2.3.3.3.

Ill



/

!

-

— ^

i

I

H*

L*

1

I

[so und da

gEhts

0.4

0.8

0.4

0.2

π η weg

0.6

1.2

0.8

I

1 L* 1

H%

I

I

RUN

ter

1.6

1.2

0.8

0.4

I Η

2

fast im rechten

0.2

2.2

1.6

1.2

0.4

0.4

1 Η 1

WIN

0.6

el

0.8

0.3

0.6

0.9

1.2

0.8

1.2

1.6

2

1.4

2.2

Figure 5.14: Simple rising accent in SaxG on phrase-final trochees (a and b), on phrasefinal words with word stress prior to the antepenultima (c and d), and on non-phrase-final words (e and f).

As for the simple fall, the accent shapes presented so far are contrasted with accent shapes realised on a phrase-final syllable in order to illustrate the effect of tonal crowding. In this case, the pitch accent and the boundary tone are associated with one and the same syllable (cf. (9)), and possible effects on the realisation of the pitch accent may be visible in the contour. Figure 5.15 illustrates two cases of phrase-final rising pitch accents. (9)

urn dieses MEHL

/N

L*H H%

'around this flour'

112

(a)

300·

r

j

150·

100·

1

200

J

Ν

1

100

I

Η H%

1

L*

1

urn dieses 0.4

300

150

L*

0.2

(b)

400

I

MEHL 0.6

jetz um den also

0.8 0.9

0.5

I I

Η

L*

hAlb

um den 1.5

I I

/

I

Η H% I

SEE 2.5

Figure 5.15: Simple rising accent on phrase-final syllables: (a) [me:l] 'flour', (b) [ze:] 'lake'.

As shown, it is impossible to distinguish between an unspecified and a high boundary tone, since the rise fills out the whole syllable. We do not find any clear transition neither towards a plateau or a further rise in phrase-final rises in our data. As we have suggested in case of SwabG simple rises (cf. chapter 4, section 4.2.3), it appears that the simple rise that is realised with a high boundary tone exhibits a final high tone that is significantly higher than any other high tone in the intonation phrase, thus the pitch range appears to be expanded. Our data presentation reveals that the accent shape of the final rise by and large does not depend on the position of the accent in the phrase. The rise is implemented with respect to the accented syllable: if postnuclear syllables follow either a final rise in pitch or a high pitch plateau arises. On phrase-final syllables the rise is implemented on that syllable. These two conditions are contrasted schematically in Figure 5.16.

(a)

(b) vo:n

να.

gan

I

I

I

L*

Η

H%

ze:

L*H H%

Figure 5.16: A schematic representation of the rising accent pattern in (a) nonphrase-final, (b) phrase-final position; shading indicates word stress.

Type (ii) of continuation concerns instances in which speakers intend to complete complex sentences. Otto von Essen has called an intonation pattern that is correlated with this type of continuation as progredient for Standard German ("weiterweisende (progrediente) Aussprüche", von Essen, 1964, 37, cf. chapter 2, section 2.3.3.3). In (10), the speaker uses two intonation phrases to express the complex sentence, which is intonationally separated at the end of the subordinate clause by final high pitch. The first intonation phrase of (10), illustrated in Figure 5.17a, is thus an example of progredient intonation.

113

(a) 250

(b)

200

•ν

200

150

100 1

100-

I

wenn die 0.2

90

Η ι

L* AUS 0.4

summt

ge 0.6

hat 0.8

0.9

ι

L* 1 und

am 0.2

04

I

Η H% I LAMM 0.6

0.7

Figure 5.17: Simple rising accent in SaxG progredient (a) and rising intonation (b).

(10)

[ wenn die AUSgesummt hat ]IP [ kAnnste vorBEI ]IP L*H H* H*L '[ if it finished humming ] [ you may go along ]'

In Figure 5.17b, we have a case of a simple rise with a high boundary tone, where the speaker projects to the next intonation phrase to complete his sentence which is illustrated in (11). (11)

[ und am LAMM ]ip [ biegst du dann wieder nach rEchts AB ]ip L*H H% L*H L*H H*L '[and at the lamb] [you are turning again to the right]'

Our data, thus, does not show that one particular intonation pattern signals a particular continuation type, yet we have a strong correlation of progredient intonation (L*H) signaling sentence or turn continuation (type II), and a simple rise with a final rise in pitch (L*H H%) signaling list intonation (type I). We have observed the same correlation for SwabG, cf. chapter 4, page 69. We turn now to simple rises in questions. Figures 5.18a, b present wh-questions, and Figures 5.18c, d yes-no-questions; Figure 5.18c is a declarative question, and Figure 5.18d is characterized by verb inversion. The tonal association of Figures 5.18a, c is given in (12). As for the declaratives, rising accents in questions show the same accent shape, i.e. a low tonal target at the beginning of the accented syllable, followed by a high target realised either in or on the postaccentual syllable. Comparing the nuclear contours of questions with the ones of declaratives presented above, we notice that questions generally are accompanied with a high boundary tone, whereas declaratives may end either in a final rise in pitch or in a pitch plateau. Thus, an intonation phrase ending in a pitch plateau is unambiguously realised on a declarative, yet a final rise is not exclusively connected to a question.

114 (a)

400300-

200150-

1 L* 1

1 Η 1 is

WAS 0

0.1

1 H% 1

0.2

da 0.3

0.4

0.5

(c)

220200-

/

f

Ν

10090-

L* 1 geh=ich da

0

Η H% 1 RUN

ter 0.3

0.1 0.2 0.3 0.4 0.5 0.6

0.6

0.9

1.2

1.5

Figure 5.18: Simple rising accent in questions.

(12)

Tonal association of the nuclear rise on a wh-question (12-a) and on a yes-no-question (12-b). a.

b.

WAS ist da

Κ

'what is there'

I

L*H H% geh ich da RUNter

I Λ

'do I go downward there'

L * H H% In SaxG, questions may also be realised with a rising nuclear pitch accent followed by a low boundary tone (Kügler, 2003a). This intonation contour raises the impression of overall falling intonation. However, we did not find any instance of a rising-falling nuclear contour in our map task data. The questions containing a low boundary tone discussed in Kügler (2003a) were found in natural everyday conversations of the same speakers who participated in the present investigation. In Kügler (2003a), this particular type of question that diverges from a kind of universally assumed canonical intonation pattern of questions (cf. Bolinger, 1978) has been analysed as expressing a speaker's expectation with regard to the information status of the answer. A low boundary tone is an indication of the speaker that he assumes the proposition of the question to be true, and this assumption is based on information gained from the previous context.

115 A similar phenomenon has been reported for Italian, where it is a falling pitch accent (and not the boundary tone) that signals that a speaker is asking for given information (Grice and Savino, 1997). The "canonical" form would be a question realised with a rising nuclear pitch accent, which signals that the speaker is asking for new information. In follow up studies, Grice and Savino (2003a,b) extend their notation of information structure in that they take the speaker's consciousness into account. Thus, a three-way distinction of information status arises: besides given and new information speakers are conscious about accessible information as well. The intonation pattern that speakers use to indicate information or confirmation questions depends on the speaker's degree of confidence in the information being asked. Basically, the simple rise is a perfect inversion of the simple fall. Both accent patterns are associated with the stressed syllable, where the change in pitch begins at the onset and ends at the offset of the stressed syllable. The rise conveys a general meaning of incompleteness and openness. This is best represented in the continuation case when a speaker has not completed his utterance. In the case of a question, the phrase is not completed or open in the sense that the speaker is querying for an answer. This kind of attribution of meaning has been reported for Standard German (Fery, 1993) as well as for German dialects (e.g. Tyrolean German, Barker, 2002). A further case of meaning that has been attributed to the rise in Standard German as well as in Tyrolean German (Fery, 1993; Barker, 2002) can also be found in SaxG: an expression of curiosity or surprise. (13) illustrates this with a passage, where Speaker A is obviously surprised over the symbol 'caravan' that Speaker Β is referring to on her map (cf. lines 2 and 4). Speaker A is still considering the previous symbol 'idyllic pond', when speaker Β refers to the symbol 'caravan'. Although Speaker A just heard about the caravan, since Speaker Β just mentioned this symbol, A gets surprised uttering "what is there?" in line 4. The meaning of surprise associated with the wh-question is also accounted for by Speaker Β when she is laughing about the symbol (cf. line 5). (13)

1

A:

2

B:

3

A:

j e t z könnten wir now we might

is [da der WOHNwagen reschts is the caravan there to the

right

[also der is [ok that is

—• 4

WAS is da L*H H'/.

(cf. Figure 5.18a)

what is there

5

B:

bei mir is so ((lachen)) by me it's so

((laughter))

Summary The simple rise contour is almost as common and frequent as the simple fall in SaxG occurring in declaratives as well as in all kinds of questions. The simple rise is a perfect inversion of the simple fall with respect to its tonal make-up and its phonetic alignment, it is composed of an accentual low tone that is realised at the beginning of the accented syllables' nucleus, followed by a high trailing tone that is realised at the end of the accented syllable in half of the cases, or in the first

116 postnuclear syllable in the other half. A schematic representation of the simple rise is given in Figure 5.13 above. We distinguish between two phonetic variants, one with a final rise in pitch and one with a high postnuclear plateau. Questions are generally ending in high pitch, i.e. with a final rise in pitch, whereas declaratives may end either in a final rise or in a high pitch plateau. The phonological form of the simple rise in terms of the tone-sequence model is L*H H%. We have argued that the accentual low tone is perceptually more prominent and shows acoustically a more invariant pattern than the high tone. Therefore we analyse the high tone as a trailing tone. A phonetic variant of the simple rise is labelled L*H expressing that the intonation phrase boundary is tonally unspecified. For declaratives we distinguish two types of continuations expressed by the rise, a list continuation and a sentence or turn continuation. We find a strong correlation between a simple rise with a high boundary tone expressing list continuation, and between a simple rise with a high plateau expressing sentence or turn continuation, yet on the basis of our data there appears no categorical distinction between the two variants of the simple rise.

5.2.4

The fall-rise

In this section, we present evidence for the fall-rise contour, which we analyse as a sequence of a falling nuclear pitch accent plus a final rise in pitch (H*L H%). We assume that the fall-rise is a base pattern of SaxG, though we identify only a few examples in our speech data, a fact that we assume is an effect of a corpus study rather than a frequency effect. From a phonological point of view, it appears reasonable to assume that the fall-rise is a base contour along with the simple fall and the simple rise (more on this issue will be discussed in section 5.2.8 below). Two reasons support this position. First, following Gussenhoven (1984), with respect to meaning, we interpret the fall-rise to convey the basic meaning of SELECTION, cf. chapter 7, which expects the contour to be a base contour. Second, in section 5.2.5 below, we analyse a similar contour as a modification of the fall-rise. If our assumption of modified contours in SaxG holds, then the fall-rise must be a base contour.

H*

L

H%

Figure 5.19: A nuclear fall-rise in SaxG schematically represented in a three syllable environment; grey shading indicates word stress.

The falling accent is similar to the one of the simple fall contour. If the nuclear falling accent is realised early in the intonation phrase, a low pitch plateau arises that extends to the postnuclear syllables until the final rise in pitch. A schematic representation of the fall-rise is given in Figure 5.19.

117

J

/

H* 1 da 0.1

0.2

0.2

0.3

0.4

0.4

be

0.8

H% 1

ginnt 0.3

0.6

L 1

0.6

mein 0.9

0.9

weg 1.2

1.5 1.6

15

0.5

2.2.6

Figure 5.20: Fall-rises in SaxG. Figure 5.20 presents examples of the fall-rise. In Figure 5.20a, a fall-rise is shown that is realised on the phrase-final syllable LAMM 'lamb'. We observe the similarity of the simple falling accent, i.e. the fall starts early in the accented syllable, and reaches a low target in the accented syllable. In this example, however, one more tone is implemented on the accented syllable, i.e. the final rise. Thus, phonologically, three tones associate with the accented syllable, e.g. (14-a). (14)

a.

e LAMM

b.

H*L H% und dann geht=s gradeaus UNten weiter

IV

'a lamb'

I /

\

H*L H% 'and then it continues straight on below' In Figures 5.20b, c and d, the fall-rise is implemented on more than one syllable, and the low postaccentual plateau becomes obvious, e.g. the tonal association of Figure 5.20d in (14-b). In these cases, the falling accent is identical in form to the one of the simple fall. The crucial difference, however, is the second high tone, which we analyse as a boundary tone. In the fall-rise, pitch rises significantly at the end of the phrase. An interpretation of the second high tone as a second trailing tone is therefore ruled out. The fall-rise contour is found in Standard German as well (Fery, 1993, 91ff), identically analysed as H*L H%. According to Fery, the Standard German fall-rise goes with questions but might also express a threat.

118 5.2.5

Modified nuclear contours

In this section, we will analyse two further nuclear contours of SaxG intonation, the rise-fall and the rise-fall-rise. We will argue that these contours share a particular intonational feature, i.e. they start with a rise. We assume that the rise at the beginning of the two contours is a result of the tonal modification L-prefixation (cf. Gussenhoven, 1984, 2004, for L-prefixation in British English). According to Gussenhoven (1984), a tonal modification affects the base contours of a language. What is crucial is that a tonal modification is identical across contours. This criterion is fulfilled here, since the two contours start with a rise, which we will analyse as an inserted low leading tone ( L + ) . The insertion is indicated by the ' + ' . Further, a modified contour should show obvious structural similarities with its base contour. This is also the case for the two contours analysed in this section, the rise-fall shares the falling nuclear accent ( H * L ) with its base contour simple fall, and the rise-fall-rise shares the whole nuclear contour shape of its base contour fall-rise (H*L H%). Figure 5.21a presents a schematic representation of the rise-fall, Figure 5.21b of the rise-fall-rise.

/ (a)

[ZZjtZZllZD

(b)

EZZ1IIZ3IZD

I

L+H*

I

L

L+H*

I

L

I

H%

Figure 5.21: A schematic representation of two modified contours in SaxG; (a) rise-fall, (b) rise-fall-rise; grey shading indicates word stress.

Phonetically, we observe a low turning point in pitch prior to the accentual fall in both contours, which results in an Fo rise on the accented syllable. This rise appears to be similar to the one of the simple rise. Perceptually, however, the rising part does not sound like that of the simple rise. Whereas the simple rise exhibits a perceptually salient low tonal target, the rise in the rise-fall / rise-fall-rise contours does not. Rather, the accentual fall appears perceptually salient, which points to a structural similarity with the simple fall and the fall-rise, respectively. With respect to the alignment of the pitch peak, the rise-fall and the rise-fall-rise exhibit the same Η alignment (around 110 to 120 ms into the vowel of the accented syllable, cf. Table 6.2 in chapter 6). The Η tone alignment differs significantly from the simple fall (cf. chapter 6). Recall, that in case of the simple fall, the Η tone is aligned near the beginning of the vowel. Phonologically, we propose that the preaccentual low tone is best understood as a low leading tone, since it is less salient perceptually than the accentual high tone. Further, since the preaccentual low tone is found in the case of the rise-fall and the risefall-rise, we assume that it is a matter of a systematic modification of base contours, i.e. L-prefixation, which has been assumed for British English (Gussenhoven, 1984, 2004). According to Gussenhoven, in English, L-prefixation results in a delayed base contour so that the inserted low tone receives accentual prominence. One might think of a similar analysis for our two contours here, i.e. the rise-fall could be interpreted

119

as L*HL, and the rise-fall-rise as L*HL H%. We reject such an anlysis for the SaxG contours since we observe a significant difference in L tone alignment between the leading low tone of the rise-fall and rise-fall-rise, and the simple rise. If we were to assume a phonological analysis of L* for the SaxG contours, we would expect the tonal alignment of the L tone to be identical (cf. the invariance hypothesis of tonal alignment, chapter 6). The observed significant difference of the accentual Η tone alignment between the rise-fall contours and the simple rise, then, is simply a matter of the insertion of a low tone, i.e. the Η tone is shifted rightwards into the accented syllable. We distinguish between the rise-fall and the rise-fall-rise. This is done analogously to the distinction between the simple rise and the fall-rise (see above). The crucial difference between the rise-fall and the rise-fall-rise is the final rise in pitch, which consistently occurs on the final syllable of the intonation phrase (cf. Figure 5.22 below). Thus, we interpret the final rise as caused by a high boundary tone. The rise-fall and the rise-fall-rise occur on declaratives, but may also be used in questions (cf. Table 2.2, chapter 2). The realisation of these contours is shown in Figure 5.22, where non-phrase-final realisations of the rise-fall are given in Figures 5.22a, c and e, and of the rise-fall-rise in Figures 5.22b, d and f. Figures 5.22a and b present two realisations on a phrase-final trochee, Figures 5.22c and d two realisations on accented syllables prior to the antepenultimate, Figures 5.22e and f, two realisations on a non phrase-final nuclear word. The autosegmental association is given in (15). (15)

a.

die reschte seite vom WEIher

b.

L+H* L da können wir unsere aufgabe WIDmen

c.

L+H* L H% 'so we can concentrate on our task' bei der trAurigen trAuerweide (.) am NONnenweiher

d.

L+H* L 'at the sad weeping willow (.) at the idyllic pond' dort steht ne TRAUerweide 'there is a weeping willow'

e.

L+H* L H% dann: würde bei mir das LAMM kommen L+H*

f.

'then the lamb would come on mine' eine DREHende Windmühle

L+H* L

H%

'the right side of the pond

L 'a rotating windmill'

120

(a)

180

(b)

280 200

150

J

Λ

- -vA 100

I I I L+ H* L ι ι I vom

WEI

κ

ν

J

I I I I H*I I

L+ her

0.4

0.2

.

100 90 [da können wir unsre aufgabe (-)

0.6

0.7

0.5

1 1

L

WID

H% men

1.5

(C)

150-

/

}

2.5

2.8

(d)

180-

r

,Λ Λ

100

J

N

\

J

/

70 L*H ι

bei der

IrAu

I*

I

rigen

ι

I

ι

I

L+H*

I

trAu er

wei

0.5

de

100 90

I

II II

L

2.5

dort steht ne 0

2.7

0.2

0.4

L

TRAU

er

0.6

0.8

1 I

H%

weide 1

1.2

(0

(e)

380

1 ι

L+ H*

I

am M O nen wei her

1.5

/

J

\

J

/\ 300

J 200 180

ι L+ ι dann: [würde bei mir das

0

0.3

0.6

ι Η" ι

LAMM 0.9

1.2

I I I Η L I I I DREH ende

ι L

I

eine

kommen 1.5

1 1

L+

1.8

0

0.2

0.4

0.6

H%

Windmühle 0.8

1

1 2

Figure 5.22: Rise-falls (a, c and e) and rise-fall-rises (b, d, and f) in SaxG. Examining the rise-falls and rise-fall-rises in Figure 5.22, the question arises whether the second low tone could be analysed as a phrase accent. The low tone appears to align at a postaccentual prominent syllable. This docking behavior of a phrase accent has been shown for different languages, in particular for Eastern European languages (Grice et al., 2000). For Standard German, the GToBI system also assumes a low phrase tone docking on postnuclear stressed syllables (cf. chapter 3, and Grice et al., 2000; Grice and Baumann, 2002). We reject an analysis of the L tone as a phrase accent, however, for three reasons. First, L tone alignment appears not to support the postnuclear stress-seeking behavior throughout; we have examples in our corpus where the L tone aligns on non-prominent syllables. Moreover, consider for instance Figure 5.22f, where the low tone appears to align with the primary as well as with the secondary stressed syllable of the phrase-final word Windmühle [ v i n d . m Y i l a ] 'windmill'. According to Grice et al. (2000, 168), the L phrase accent in Standard German, has a "degree of optionality as to which lexical stress the elbow [i.e. the

121 low turning point, F.K.] synchronises with". This would thus hold for SaxG as well. The general question arises whether this degree of optionality might be taken as evidence for the claim that the L tone is independent from the nuclear peak. Prom an inspection of our data, we doubt the variation being found with respect to the alignment of the low tone might lead to an interpretation of a phrase accent, yet this remains impressionistic, and we call for a systematic controlled production experiment to test the status of the L tone. Second, we reject the analysis of the L tone as a phrase accent analogous to the simple fall (and the fall-rise), where we obviously did not find postnuclear stressseeking behavior of the low tone. Instead, in about 50 % of the cases, the low tone is realised within the accented syllable. Based on the idea of modification, i.e. a structural similarity between a modified and a base contours we conclude that the L tone in this contour is better understood as a low trailing tone. Third, as we will argue in chapter 7, we reject the analysis of phrase tones in SaxG since the theory of intonational meaning put forward by Pierrehumbert and Hirschberg (1990) that assumes phrase accents as belonging to the tonal inventory of languages fails to capture this particular SaxG intonation pattern in terms of a compositional analysis. We will now briefly present phrase-final realisations of the rise-fall (cf. Figure 5.23). We observe that the shape of the contour is compressed to the accented syllable. In final position, the peak is slightly shifted to the left so that the falling part of the accent receives enough space to be realised. The autosegmental association of phrasefinal rise-falls is given in (16), where it is shown that three tonal targets are associated with the phrase-final accented syllable. (16)

a.

a JUNger (-) juweLIER

'a jung (-) jeweller'

L+H* L b.

en L A M M

'a lamb'

/V

L+H*L

(a)

250200-

150-

100-

J

V

Η

I I L+ H* I I a

(b)

>

I L I

JUN 0.3

/\ I L+ I

Η

ger 0.6

I H* I

ν

1.2

>

I L I

LIER

ju we 0.9

f\\



\ I L+ I

en 1.5 1.6

Figure 5.23: Rise-fall in SaxG on phrase-final accented syllable.

0.2

I H* I

I L I

LAMM 0.4

0.60.7

122

In our corpus, we did not find any instance of a rise-fall-rise realised on a phrasefinal syllable, though we expect these patterns to occur. It might be the case that speakers tend to move an accented monosyllabic word from the phrase boundary to avoid tonal crowding, since a rising-falling-rising nuclear contour takes a highly marked effect phrase-finally (cf. Fery, 1993, 91, who makes this point in case of a phrase-final fall-rise pattern, i.e. H*L H%, in Standard German). It may simply be physiologically unrealistic to implement four tonal targets on one single syllable. This must be shown in a future experimental study, however. Finally, we present examples with the rise-fall and the rise-fall-rise realised on questions. This is shown in Figure 5.24a for an echo question and for a yes-no-question in Figure 5.24b with the autosegmental association in (17).

350-

(a)

/

300-

(b)

X ;

•/

\



·

y

·

200·

150

1 H* 1

1

L+ I

[keen

NON 0.2

nen 0.4

1 L 1 wei

her

0.6

0.8

I L+ I

0.4

0.6

1 H% 1

L

WOHN

hast du en 0.2

I I H* I I

wa 0.8

gen 1.2

Figure 5.24: Rising-falling accent in SaxG on questions.

(17)

a.

keen NONnenweiher

b.

L+H* L hast du en WOHNwagen

'no idyllic pond'

Λ \

/A

L+H*

I 1

'do you have a caravan'

L H%

With regard to linguistic function and meaning of the rise-fall and the rise-fall-rise, we suggest that these intonation contours correlate well with the expression of narrow focus. In section 4.2.8 below, we will analyse this correlation for the rise-fall. In addition, our analysis of intonational meaning in chapter 7 attempts to show that the basic meaning of the unmodified contours of the rise-fall and the rise-fall-rise, i.e. the simple fall and the fall-rise, is supplemented by the meaning expressed by means of L-prefixation, i.e. L-prefixation adds a nuance of 'significance' to the basic meaning, which correlates well with the expression of narrow focus. Summary The rise-fall and contours of the base contours simple exhibit a low tonal target near the significantly from the accentual low

rise-fall-rise contours are analysed as modified fall and fall-rise, respectively. Both contours

beginning of the accented syllable that differs tone of the simple rise. Since the preaccentual

123

low tone is phonetically identical in both contours and less prominent than the high tone, we analyse it as a low leading tone that associates with the accented syllable. Accordingly, the nuclear falling accent of the base contours is shifted rightwards into the accented syllable, yet the basic contour shape remains, which allows us to assume that the L tone is a result of a tonal modification of L-prefixation. The crucial difference between the rise-fall and the rise-fall-rise contours is in the boundary tone, i.e. the rise-fall-rise shows a significant rise at the phrase-final syllable. A schematic representation of the two contours is given in Figure 5.21 above. The phonological form of the two contours is reminiscent of the form of the base contours; the rise-fall has the form L+H*L, the rise-fall-rise L+H*L H%. The ' + ' between the leading low tone and the accentual high tone indicates L-prefixation. The distribution of the two contours is not restricted to sentence types, i.e. it occurs on declaratives as well as on all types of questions. Both contours add the same kind of meaning to the meaning of the corresponding base contour. In terms of Gussenhoven (1984), we assume that L-prefixation adds a nuance of 'significance'. The rise-fall is a means of expressing narrow focus.

5.2.6

Boundary specifications

In SaxG intonation, we identify a high boundary tone H%. In addition, phrase boundaries may be tonally unspecified. This analysis follows the approach taken by Fery (1993) and Grabe (1998a) (cf. chapter 3). The high boundary tone accounts for the obvious occurrence of final rises in pitch that are associated with the edge of an intonation phrase and not with a prominent syllable. On the contrary, we did not observe a similar final falling pitch movement. To account for low pitch at an intonation phrase boundary, we assume with Fery (1993) that the trailing low tone spreads until the intonation phrase boundary. If an accent occurs early in an intonation phrase, pitch may fall slightly gradually until the end of the phrase, which we assume to be an effect of declination. The SaxG boundary specifications are identical to Standard German (Fery, 1993), yet differ from SwabG as discussed in chapter 4, section 4.2.7.

5.2.7

Focus

In this section, we will briefly analyse the focus structure of two examples to illustrate a case of narrow and a case of broad focus. We show that narrow focus is expressed by means of a rise-fall (L+H*L), whereas broad focus by means of a simple fall (H*L). In case of the rise-fall, we analyse in about 90 % of the cases a narrow focus. Consider the case of narrow focus in (18) given in a broader context. The target sentence is indicated by an arrow. Narrow focus is on the word BLAUe 'blue'. According to our discussion about focus in chapter 3, section 3.2.4 and in line with Ladd (1980), we analyse narrow focus if the focus of a sentence is on a particular constituent of the sentence, which is indeed the case in (18). From the context it is obvious that the style of the arena is focused, that is, speaker A corrects herself in changing the adjective from 'new' to 'blue'. In (18-b), we present the focus analysis

124

with square brackets indicating the domain of focus. The contour which is realised on the sentence in (18-b) is the rise-fall. (18)

Narrow focus in SaxG a.

1

A:

käme jetz eine nEue aREna would come a new arena

2

B:

bei mir kommt ne blAue aREna here comes a blue arena

—• 3

A:

dann: nimmste die BLAUe arena then take the blue arena

b. B:

dann nimmste die [ BLAUe ]poc arena L+H*L

In (19), we present a context with a target sentence in broad focus, i.e. the whole sentence is focused as indicated by square brackets. Prom the context, it is obvious that no particular constituent of the target sentence is in focus. Speaker A is going on with the instructions of how to draw the route on the map. None of the words of the target sentence has been mentioned before, and no particular constituent of the sentence is narrow focused. In this case, according to Ladd (1980), default accentuation should be, and indeed is visible (cf. the tonal representation in (19-b)). One might object that the information of which symbol is coming next is not very new or unexpected. Thus, an interpretation of narrow focus might be possible, where the item 'Iranian army' is narrow focused in contrast to all other symbols. However, the participants are at a phase in the map task, where they already have recognized that the maps differ from one another. Thus, which particular symbol is coming next and at which location is crucial to the task, and presents entirely new information. (19)

Broad focus in SaxG a.

1

A:

so haste den jetz praktisch hier so eingekreist

2

B:

genau (.)

basically you have it circled exactly 3

das hab [isch jetz gemacht that I have done now

4

A:

5

B:

—> 6

A:

[nach unten jetz geht=s wieder nach link[s downwards now it's going again to the left [ja hab isch och yes I do have it also und dA kommste bei der irAnischen arMEE [(.) and then you are passing the Iranian

7

B:

army

[ach du grosser gott was wollen denn die hier oh dear what do they want here

b.

[ und da kommste bei der iranischen Armee vorbei ]FOC L*H H*L

vorbei

125 If we consider the tonal analysis of (18-b) and (19-b), we observe that narrow focus is expressed by means of a rise-fall, broad focus by means of a simple fall. We observe a strong correlation between focus type and accent type in our data. We assume, therefore, that speakers of SaxG differentiate intonationally between broad and narrow focus. The fact that the tonal structure differs with respect to focus type has been reported for different languages, e.g. Hayes and Lahiri (1991) for Bengali, Zybatov and Mehlhorn (2000) and Alter (1997a,b) for Russian, and Frota (2000) for European Portuguese. For Standard German, a tonal difference between a falling accent (H* L-) and a rising-falling accent ( L + H * L-) has also been claimed in the GToBI system (Grice and Baumann, 2002), but not in the approach taken by Fery (1993) and Grabe (1998a). In GToBI, the rising-falling contour has been ascribed to contrastive focus (cf. the discussion about the G T o B I system in chapter 3). Recently, in his analysis of Standard German, Peters (2006) also claims that broad and narrow focus are signaled by means of distinct accentuation. For SwabG, we have also found this distinction (cf. chapter 4, section 4.2.8). Interestingly, the contour shape of SwabG narrow focus is fairly similar to the one of SaxG, i.e. SwabG L*HL, and SaxG L + H * L , yet the phonological differences are obvious. Nevertheless, in both dialects, it is a modified contour that expresses narrow focus.

5.2.8

The phonology of Upper Saxon German nuclear contours

This section intends to summarize our previous analysis of nuclear contours in SaxG. We propose that the tonal system consists of three nuclear base contours that may be modified by means of L-prefixation. We observe, however, only two modified contours in our corpus. The missing contour might be fairly rare so that we did not record it. Our analysis of SaxG nuclear contours suggests that we distinguish between three basic contours, the simple fall (H*L), the simple rise (L*H H%) and the fall-rise (H*L H%). In addition, we observe two further nuclear contours, the rise-fall ( L + H * L ) and the rise-fall-rise ( L + H * L H%). With respect to intonational form, the above analysis has shown that each contour type exhibits its own invariant shape. T h e realisation of the accent shape of a nuclear contour does not depend on the accent position in the intonation phrase. Accents realised early in an intonation phrase have the same form as the ones produced late. In particular, this holds for accents realised on phrase-final accented syllables as compared to non-phrase-final realisations. This is different from Standard German, where Grabe (1998a) showed that the realisation of the simple falling accent depends on its position in the phrase (cf. also Goldbeck and Sendlmeier, 1988). Phrase-final falls align the high tone at the beginning of the syllable nucleus, whereas non-final falls align the high tone at the end of the stressed syllables' rhyme (Grabe, 1998a, 75, Figure 10, and Figure 3.5 in chapter 3). Thus, for Standard German Grabe assumes differences in intonational form to be related to the same phonological category underlyingly. However, for SaxG we do not observe this kind of variation, which is supported by the phonetic analyses in chapter 6, where we show that tonal alignment of each of these three accent categories is fairly invariant.

126

For the simple fall and simple rise, we observe that the second tone of a bitonal pitch accent undergoes spreading. In line with the theory of intonational phonology, we take the behavior of the second tone as an indicator that differentiates between the prominence relations of the two tones in bitonal pitch accents. Tones that undergo spreading are assumed to be trailing tones. With respect to tonal alignment, the accentual tone in the simple fall and simple rise aligns invariantly in relation to the accented syllables' nucleus. The trailing tones, however, are half of the time realised in the accented syllable, i.e. near the end, and in the other half of the cases on the first postnuclear syllable. The realisation of the second tone is thus in relation to the accentual tone which suggest an interpretation of a trailing tone. The perceptual impression is in accordance with these phonetic facts, that is, the accentual tone is perceptually more salient, thus more prominent than the trailing tone. For the fall-rise, we assume the same generalizations to hold, yet, we have not a sufficient number of examples to carry out statistical measurements. The crucial difference of this contour compared to the simple fall is the final rise in pitch. Since the rise in pitch generally occurs in the phrase-final syllable and is not related to metrically strong, i.e. prominent syllables, we analyse the rise as a high boundary tone. For the simple rise, we identify two different contours in our corpus, a simple rise with final rise in pitch (L*H H%) and one with a high pitch plateau (L*H). Our data do not suggest a difference in function or meaning between the two. Although we distinguish between two types of continuation in case of declaratives, we cannot relate a particular simple rise contour to a particular continuation type. We have found a strong correlation, though, that the simple rise with high plateau appears to project only to an adjacent intonation phrase, and the simple rise with a final rise in pitch projects to a larger domain. This distinction appears not to be categorical, however. Therefore, we treat these two contours as phonetic variants of the simple rise. For SwabG, we have found the same phonetic variants. Apart from the three base contours, we identify two further nuclear contours, the rise-fall and the rise-fall-rise, which both share a preaccentual low tonal target. Based on this similarity, we analyse these two contours as modified versions of base contours. The two contours, in fact, show striking structural similarities to the simple fall and the fall-rise. We analyse the simple fall and the simple rise as base contours of SaxG, an assumption based primarily on their frequency. For the simple fall, we have shown that this particular accent is used as default accentuation, which is a further hint to analyse it as a base contour. Finally, in chapter 7, our attempt to analyse the meaning conveyed by the simple fall and simple rise further supports our assumption of base contours. For the third base contour, the fall-rise, we have no frequency effects to rely on our assumption of analysing this contour as a base contour, yet given our analysis of modified contours, it appears reasonable to assume the fall-rise as a base contour due to its structural similarity to the rise-fall-rise. Based on the analysis of base contours, we analyse the remaining two nuclear contours as modified contours, mainly due to their structural similarity to the base contours. A further criterion for a modified contour is fulfilled, i.e. the modification is structurally identical across contours, in our case L-prefixation.

127 We have rejected an alternative analysis which would be based on the assumption of a phrase accent (Grice et al., 2000), which is included in the GToBI system (e.g. Grice and Baumann, 2002). The nature of the phrase accent in German has been experimentally investigated by Benzmüller and Grice (1998) who showed that the shape of the fall in a falling accent depends on the number of unstressed syllables that follow the nuclear accented one, i.e. the fall aligns with a postnuclear prominent syllable. Three major questions arise here. First, we doubt that these findings can be generalized since the analysis is only based on two speakers. Second, Benzmüller and Grice (1998) found a certain amount of variation in their data (see above, page 121). Thus, the evidence for a phrase accent in German is very weak. The third question is more general with respect to our data: According to (Grice et al., 2000), a phrase accent is an edge tone that is secondarily associated with a postnuclear prominent syllable. Our SaxG data, however, does not support an analysis of the postaccentual low tone as being secondarily associated with a postnuclear prominent syllable, cf. in particular the rise-fall and rise-fall-rise. Instead, our data suggests an analysis that the postaccentual low tone belongs to the accent since it is perceptually salient connected with the accent and it is realised close to the accent. If analysed as an edge tone the question would arise as to why an underlying edge tone causes the impression of an accentual tone on the surface. However, leaving these objections aside and taking the analysis of a phrase accent seriously, we should find a stress-seeking behavior of a low tone in case of the rise-fall and rise-fall-rise (see above). Consequently, if analysing the second L tone as a phrase accent, we should also analyse the L tone of a simple fall as a phrase accent. Yet, the simple fall L tone does not show a stress-seeking behavior at all. Thus an analysis of the L tone in a simple fall as a phrase accent could only be based on an analogy to the rise-fall contour. We doubt, however, such an analogy and would prefer it the other way round, that is, given the high frequency of the simple fall and its status as the default accentuation, we assume that the analysis of a trailing L tone should carry over to the rise-fall contour, where we analyse the second L tone, though with a tendency of being stress-seeking, as a trailing tone as well. In sum, we have different reasons to reject an analysis of a phrase accent. Rather, we suggest that the complexity of nuclear pitch accents, e.g. L + H * L , arises by means of L-prefixation keeping the structural similarity to the base contours.

5.3

Prenuclear accents

This section is concerned with prenuclear accents in SaxG, i.e. any accent in an intonation phrase but the final. In prenuclear position, falling as well as rising accents may occur. Each of the prenuclear accent shapes is examined separately, section 4.3.1 examines falling, and section 4.3.2 rising accents, and we argue that prenuclear rising accents function as topic marker.

128

5.3.1

Prenuclear falling accents

Prenuclear falling accents may occur in combination with nuclear falling or rising accents. Figures 5.25a, b show a sequence of two falling accents. The accent shape of prenuclear falling accents is identical to the one of nuclear accents. In Figure 5.25a, b, the high and the low tone align on the accented syllable, respectively, and the high tone is realised early. An example of a sequence of a prenuclear falling and a nuclear rising accent can be seen in Figure 5.14f above. Again, the accent shape of the prenuclear falling accent is identical to the nuclear one.

\ \— ν 1 H* I kEe

ι !H* 1 ne 0.2

TRAU 0.4

I L 1 er 0.6

weide 0.8 0.9

das

\

\

I I Η* L I I

II HI. II

kOmmt 0.3

\

jetzt

ERST

0.6

0.9

noch 1.2

1.4

Figure 5.25: Prenuclear accents; (a, b) a sequence of two falling accents; (c) a partially linked prenuclear falling accent; (d) a completely linked prenuclear falling accent.

We identify two further cases of prenuclear high tones, one where the prenuclear accent is partially linked (Figure 5.25c), and a second where the prenuclear accent is completely linked (Figure 5.25d). In line with Gussenhoven (1984), and for German with Fery (1993) and Grabe (1998a), we assume that these prenuclear accents are realisational variants of an underlying falling accent. For the discussion about tonelinking see chapter 3, page 32, and for tonal linking rules see (7) on page 33. The completely linked version in Figure 5.25d represents a hat pattern, more precisely following Fery (1993, 149) it is hat pattern 1. The second hat pattern arises in a sequence of a prenuclear rising and nuclear falling accent (see next section). The partially linked prenuclear falling accent in Figure 5.25c exhibits a delayed fall, that is, the pitch stays high on the accented syllable (cf. the pitch on the ac-

129

cented syllable kömmt 'comes'). The accentual fall starts first on the post-accentual unstressed syllable if there is one. In case of a completely linked falling accent, the low tone is deleted, as can be seen in Figure 5.25d. Interpolation occurs between the first high tone, which is associated with the syllable sie 'she' and the next high tone, which is associated with the stressed syllable of the verb schildert 'describe' (cf. the tonal association in (20)). Following Grabe (1998a), we use the label ' > ' to indicate that it is not a monotonal high pitch accent (H*) but a realisational variant of an underlying falling accent. (20)

wie sie mir das jetzt SCHILdert H*>

5.3.2

'as she is describing it to me'

!H*L

Prenuclear rising accents

A prenuclear rise may occur with any of the three nuclear accents. The most frequent pattern of a declarative sentence is the hat pattern, a tonal sequence of a prenuclear rising and a nuclear falling accent; following Fery (1993, 149) this tonal sequence is categorized as hat pattern 2. Two examples of the hat pattern are given in Figure 5.26. (a)

350300

(b)

Ν

200

\

100 90

1

1

ι

L*H

Η*

L

I

dir

alos=bei 0.3

V

ι

geht=s zur 0.6

TRAU 0.9

I

L* I

ι

er

weide 1.2

ΘΠ

1.4

I I

I

Η

wOhn 0.2

Η* wagen 0.4

I

HAB 0.6

I

L I

=ich

0.8

Figure 5.26: A tone sequence of a prenuclear rising and a nuclear falling accent; ('hat pattern 2', cf. Fery, 1993, 149).

If we compare the accent realisation of the prenuclear rise with the nuclear one, we observe no difference. Consider in particular the prenuclear realisation on WOHNwagen 'caravan' in Figure 5.26b, and compare it to the nuclear realisations of this word (cf. Figure 5.5). The prenuclear accent realisation is identical to the nuclear one, each tone aligns with a separate syllable, the stared tone with the accented one. See (21) for a tonal association with the text.

130 (21)

en wOhnwagen H A B = i c h

I/

I/

L*H H*L Ί do have a caravan'

A tonal sequence of two rising accents is shown in Figure 5.27. We observe interpolation of pitch from the first pitch peak to the second low tone. We do not observe upstep of the accentual low as an effect of the first rising accent. Fery (1993, 88) mentions that upstep may optionally occur in a sequence of two rising accents in Standard German. See (22) for a tonal association. (a)

§§8

/

200

Ν

150

J

100·

ι I L* Η ι I urn rUn

1 1 L' Η 1 1 die blühenden Η

dest

0.5

(b)

180

ι ι L* Η ι ι BLU

A

150

\

J ι Η% I men 2.52.6

1.5

so 0.3

0

1 L* I wie die Ii BEL

1 I L* Η ι ι brEit

100

J

-N

0.9

0.6

1.2

/ Η I le 1.5 \

Figure 5.27: A tone sequence of two rising accents. (22)

so brEit wie die liBElle L*H L*H 'that broad as the dragon fly'

The prenuclear rise may also occur in combination with the nuclear rise-fall and risefall-rise. Examples for the latter case are given Figure 5.28. (23) presents the tonal association of Figure 5.28a. (a)

180

f

' X

0

·. •

ι I L* Η ι I es geht 10s 0.3

am 0.6

ι I 1 1 L+ H* L H% ι I 1 1 NON nen weiher 0.9

1.2

1.5

70)

0

·

/

V*

100

100

70·

1

da mAch =isch jetz ä 0

Ν

0.3

strich

0.6

\

\

Φ

1 !H* 1 zum 0.9

WOHN

wagen 1.2

hin 1.5

1.7

Figure 5.29: A tone sequence of more than two accents.

(24)

da mAch=isch jetzt ä strich zum WOHNwagen hin L*H

H*>

!H*L

'there I'll make a line to the caravan'

With respect to tonal linking, we do not observe any partially or completely linked rising accents. However, we do not assume that they must not occur. Consider a partially linked rise and fall, for instance. In Standard German, this pattern expresses a meaning of 'astonishment' (Fery, 1993, 127f). Assume that this is true in SaxG as well. The reason why we did not find any example of this pattern, then, might be that the speakers simply did not express any kind of astonishment. However, we could expect astonishment from the speakers wondering about the map task. We assume though, that this pattern, as well as the other linking patterns, might be elicitable in interviewing a cooperative speaker in an experimental setting. We will now briefly turn to functional aspects of the prenuclear rising accent. We assume that, as for Standard German (cf. Fery, 1993), a prenuclear rising accent functions as a topic accent. The rise indicates that 'more is to come', thus conveying the general meaning of openness. The topic function of the prenuclear rise is comparable to the fall-rise in English, as Fery (1993, 145ff) has shown in detail. Fery (1993, 149) concludes that "the choice of a rising tone in a prenuclear position is not arbitrary, but is governed by different discourse structures and semantic factors." This is also true for SaxG. Consider the context in (25). (25)

1

A:

bis ( - ) until

1

auf deiner rEchten seite ein WOHNwagen kommt a caravan appears on your right side

3

B:

der is bei mir weiter U N t e n by me that is below

132

5

Β:

un=da bin ich auch in der nähe ν also ν (.) and there I am next to ν so ν (.) am LAMM at the lamb gut also wOhnwagen heißt jetzt LAMM L*+H !H*+L ok caravan is called lamb now

6 —> 7

Α:

In (25), wOhnwagen 'caravan' is a given constituent in the discourse (background material), and functions in line 7 as a topic. The speakers' intention is to say something about the caravan, namely that the symbol they are talking about is not a caravan but a lamb. The topic is realised with a prenuclear rising accent (cf. Figure 5.8b above). A further illustration of a topic is given in (26). Here, the topic of the sentence is symbOl 'symbol', i.e. the item which the speaker is going to talk about. RECHTS 'to the right', then, is the focus of the sentence, i.e. the item which is related to the topic. In this example, it is a case of broad focus, the information status of the sentence is completely new. The pitch track of (26) is given in Figure 5.30.

-

symbOl ^

HECHTS™

du

L*H

dir

W e

S y m b o l to

_

righ,

!H*L

(b)

250200-

" "

-

Λ V

100 90

0

11 L* Η 11 das sym BOL 0.3

0.6

ι L ι RECHTS von dir

Η* lässt du 0.9

1.2

1.5

1.7

Figure 5.30: Topic accent in S a x G .

5.4

Upper Saxon German dialectal characteristics

The discussion of previous studies of SaxG intonation in section 5.1 has shown that contradictory claims about SaxG accentuation have been made. Some researchers claimed that SaxG is characterized by low pitched accents in contrast to Standard German (von Essen, 1940; Klinghardt, 1925; Waiblinger, 1925). Other analysed SaxG accents exhibiting high pitch (Gericke, 1963). Our data suggests that both views are correct.

133 We can only speculate on the contradictory accounts of SaxG intonation, yet we might think of two possible explanations. The first explanation would assume t h a t one and the same accent has been interpreted differently. Consider the rise-fall, which we analyse as L+H*L. This accent aligns the leading low and the accentual high tone on the stressed syllable (cf. chapter 6). The leading low tone might have caused an interpretation t h a t the accented syllable is low. At least von Essen (1940) refers to a low accented syllable with a following steep rise to a peak. His description of the shape of the accent appears to be similar to the accent shape t h a t we analyse as L+H*L. At the same time, the accentual peak might have been crucial for Gericke (1963) stating t h a t SaxG accents have characteristically high pitch. An alternative explanation would assume t h a t different accents have been considered. While von Essen (1940) certainly refers to the rise-fall, Gericke (1963) might have mainly found simple falls. If this is the case, the contradicting claims may be an effect of the speech d a t a t h a t have been investigated (single sentences for von Essen vs. monologues for Gericke). Even if we cannot decide which one of the two explanations might be the right one, neither of them does contradict our analysis. Rather, our comprehensive analysis of SaxG intonation above provides a coherent system t h a t reconciles earlier approaches. Another dialectal feature of SaxG intonation, i.e. the Fallbogen, has been described by Gericke (1963). Selting (2002a) and Gilles (2005) have reported this pattern for SaxG as well, however for Dresden SaxG, an urban variety geographically and dialectologically closely related to Leipzig. In terms of intonational phonology, we may interpret this pattern as a sequence of a high pitch accent and a high phrase tone. However, we did not find any instance of a high phrase accent in our corpus. In addition, we reject the analysis of phrase accents for SaxG (as well as for SwabG). Since we do not find any instance of this particular dialect pattern, we cannot conclude any convincing solution to account for t h a t pattern. The reason why we did not found this intonation pattern in our corpus may be related to particular contextual or situational factors t h a t we did not elicit in map task speech. This can well be the case if Selting is correct analysing the Fallbogen in contexts where speakers are rather emotionally involved (Selting, 2002a, 25). If this holds, the question arises whether the Fallbogen can be interpreted as a paralinguistic effect, i.e. an emphatic accent. On the other hand, one might interpret the postaccentual pitch curve described by Gericke (1963) as a dialect-specific interpolation mechanism between the accentual high tone and the trailing low tone. Assuming this, we simply ignore a second high tonal target. Though, since we do not have any d a t a to prove this hypothesis, the interpretation of the Fallbogen has to be left open here. The intonational properties t h a t we have analysed for SaxG differ in one aspect from Standard German, i.e. in the assumption of modified contours. The modification by L-prefixation results in a complex and thus highly marked tritonal pitch accent, L-|-H*L. In the account of Fery (1993), a rise-fall is also assumed (L*HL), yet its distribution is very restricted. See our discussion of the Standard German rise-fall in relation to the SwabG rise-fall in chapter 4, pages 49 and 91. If we compare the SaxG rise-fall with the modified SwabG rise-fall, we observe striking similarities. Both are modified, and it is a matter of an identical modification, i.e. L-affixation. The dialects differ with respect to the side where the low tone is affixed, yet the mechanism

134 is the same. Additionally, the function expressed by means of an L-prefixed rise-fall is narrow focus in both dialects. Although the general accent shape is by and large identical, the phonological differences are obvious (SaxG L + H * L vs. SwabG L*HL).

5.5

Summary

This chapter examines intonational patterns of Upper Saxon German (SaxG). Contradictory claims have been made in previous research on SaxG intonation with respect to accentuation, i.e. low vs. high accents in a neutral accentuation. Our analysis confirms both claims, and we argue that researchers have considered different accents or different aspects of an accent. Prom our analysis, we propose three distinct nuclear base contours for SaxG which we refer to as the simple fall (H*L), the simple rise (L*H H%) and the fall-rise (H*L H%). In addition, we analyse two further nuclear contours as modified contours, i.e. the rise-fall ( L + H * L ) and the rise-fall-rise ( L + H * L H%). The modification arises by means of L-prefixation, a term that has been proposed by Gussenhoven and Rietveld (1992) and Gussenhoven (2004) to account for British English modified contours. In case of a modified rise-fall / rise-fall-rise, the realisation of the preaccentual low tone is similar in both contours, a phonetic fact that supports the analysis in terms of a modification. The preaccentual low tone, however, aligns significantly different from the accentual low tone of the simple rise, which leads us to analyse the former as a leading tone. A third modified contour, a low simple rise ( L + L * H H%), might theoretically occur in SaxG, yet we find no empirical evidence for this pattern. The contour is predicted to occur in SaxG, since a tonal modification is assumed to apply to all nuclear base contours (cf. Gussenhoven, 1984). With regard to intonation phrase boundary specifications, we assume only a high (H%) boundary tone. In case of the fall-rise and rise-fall-rise, we observe a final rise in pitch that is connected to the right edge of an intonation phrase. Therefore, we analyse the final rise in pitch as a high boundary tone. Accordingly, the simple rise may also end in a final rise, though this pattern generally shows global rising pitch due to the rising pitch accent. In the other contours, we assume with Fery (1993) that the trailing tone of the nuclear accent spreads until the intonation phrase boundary, the boundary being tonally unspecified. In prenuclear position, falling as well as rising accents occur. For prenuclear accents, tonal linking results in accentual surface variation of underlyingly bitonal accents. The tonal grammar of SwabG is summarized in (27). (27)

A tonal grammar of Upper Saxon German.

135 Considering the information structural function of intonation, we observe that broad focus is expressed by means of the nuclear contour simple fall ( H * L ) , narrow focus by the modified simple fall ( L + H * L ) and topic by a prenuclear rising pitch accent.

6

Realisational differences between Swabian and Upper Saxon German

6.1

Introduction

According Ladd (1996, 199ff) languages can differ internationally in four different ways, i.e. semantically, systemically, phonotactically, a n d / o r realisationally (cf. also chapter 3, section 3.3). This chapter is concerned with realisational differences of intonation, i.e. a different phonetic implementation of an identical phonological category. In particular, we are concerned with two aspects of realisational differences, tonal alignment and pitch excursion. 1 W i t h respect to tonal alignment, several different studies have shown t h a t tones are anchored at certain parts of the segmental string, and t h a t this anchoring is fairly invariant for a given tonal category (cf. section 6.1.1). We might call this the invariance hypothesis of tonal alignment, which predicts t h a t the specific alignment of tones determines which tonal category is involved (e.g. Pierrehumbert and Steele, 1989, for the distinction between English L*H and LH* pitch accents). A given accent category may also differ in its specific alignment across dialects (e.g. Atterer and Ladd, 2004, for L*H accents in Northern and Southern German). W i t h respect to pitch excursion, very little is known about its particular role as a dialect-, and language-specific cue. In a comparison of falling accents across German varieties Gilles (2005, 164ff.) established t h a t this parameter differs between the dialects. Thus, excursion of an accentual fall appears to be a dialect-specific phonetic cue. A brief overview of studies concerning pitch excursion is given in section 6.1.2. The aim of this chapter is twofold. The first part (section 6.2) is concerned with an analysis of tonal alignment in order to present phonetic evidence t h a t supports our previous phonological analysis of SwabG and SaxG intonational patterns. In the second part of this chapter (section 6.3) we propose a two-dimensional pitch accent model t h a t quantifies the variation of individual tonal categories cross-dialectally. Based on our alignment measurements, we combine these with an analysis of pitch excursion sizes. We conclude t h a t these two factors appear to play an important role in capturing intonational variation across dialects. Prior to the analysis, sections 6.1.1 and 6.1.2 introduce the relevant aspects of tonal alignment and pitch excursion.

1

Another realisational aspect concerns end-of-phrase-effects, truncation or compression, which we leave aside here in order to concentrate on the realisation of pitch accents on fully voiced words (for an overview see chapter 3, page 42). Kiigler (2004b) showed that accents in SwabG are compressed.

137 6.1.1

The horizontal level of accent realisation

Accentual timing has been relevant for the theory of intonational phonology from the very beginning. Bruce (1977) showed that the distinction between Accent I and Accent II words in Swedish is based on a timing difference: The word accent fall starts prior to the accented syllable in Accent I words, but within the accented syllable in Accent II words. Accordingly, the timing of the tonal correlate of sentence accent (i.e. focus accent) depends on accent type: The focal high tone aligns earlier for Accent I than for Accent II words (Bruce, 1977, 50). The alignment of tones has been studied from several different angles and with different aims. First, and probably most importantly, from a theoretical point of view, the assumption of the existence of tonal targets within intonation was proven (Silverman and Pierrehumbert, 1990; Arvaniti et al., 1998, 2000; Ladd et al., 1999, 2000; Dilley et al., 2005). This aspect refers to the so-called 'levels vs. configurations' debate; for overviews see Bolinger (1951), Ladd (1980, 9ff), Pierrehumbert (1980, 28ff), and Ladd (1996, 59ff). The levels approach to intonation predicts that the turning points of an accent are aligned invariantly at a certain point in the segmental string of an utterance (invariance hypothesis). Such an anchor point might, for instance, occur at the left edge of the stressed syllable's onset or the beginning of the stressed syllable's nucleus. If the segmental duration between tonal anchor points varies, i.e. in case of short vs. long stressed syllables, the slope and the distance of the accentual movement between the two turning points would be a function of the duration between the anchor points. The configurational approach (e.g. Bolinger, 1986) favors a model of intonation in which F0 movements are crucial to the linguistic description of intonation. The start and end points of a movement correspond to the Η and L tones in the levels approach, but a given movement is expected to be constant in slope and duration. The prediction would be that no matter the segmental context, a movement always goes up or down in a given constant slope for a given constant duration. Thus, in contradiction to the levels approach, the configurational approach predicts that Η and L tones will vary-for instance as a function of segmental context or speech rate-since the movement itself appears as the phonetic manifestation of the pitch accent. A closely related enterprise has been to test the validity of a given phonological category described for languages by means of their phonetic alignment (Pierrehumbert and Steele, 1989; Kohler, 1991b,a; Arvaniti et al., 1998; Ladd and Schepman, 2003). In the 1990s a particular interest in the phonetics of intonation arose from a point of view of speech technology, where synthesized speech sounded unnatural, but was improved by implementing phonetic details of pitch accent realisation (Rietveld and Gussenhoven, 1995; Prieto et al., 1995). And finally, a growing body of single language description can be observed focusing both on detailed interpretation of phonological tonal categories (e.g. Gilles, 2005; Kügler, 2004b) and variation between language varieties (Atterer and Ladd, 2004; Willis, 2003). Across these theoretical interests, different aspects of accents and different effects on the realisation of accents have been studied: - Prenuclear accents (Arvaniti et al., 1998, 2000; Silverman and Pierrehumbert, 1990; Ladd et al., 1999, 2000; Atterer and Ladd, 2004)

138 - Nuclear accents (Rietveld and Gussenhoven, 1995; Gilles, 2005; Schepman et al., 2006) - Accentual peaks (Silverman and Pierrehumbert, 1990; House and Wichman, 1996) - Accentual lows (Ladd and Schepman, 2003; Dilley et al., 2005) - Accentual peaks and lows (Arvaniti et al., 1998, 2000; Ladd et al., 1999, 2000) - Contextual effects such as prosodic structure (Prieto et al., 1995) - Syllable structure (Rietveld and Gussenhoven, 1995; Prieto et al., 1995; Ladd et al., 2000) - Discourse structure (House and Wichman, 1996; Wichman et al., 2000) The conclusion of previous studies on tonal alignment is that tonal targets are aligned independently of each other (in particular Dilley et al., 2005) and with respect to segmental landmarks in or around the accented syllable. That is, duration and slope of movement vary as the segmental makeup of stressed syllables varies, but the individual tones are invariantly anchored with respect to a certain location on the segmental string (invariance hypothesis). The existence of tonal targets, the level view of intonation, has thus been corroborated. Moreover, pitch accents appear to be aligned language specifically. Accordingly, anchor points of an identical tonal category differ between languages.

6.1.2

The vertical level of accent realisation

Concerning the vertical level of accent realisation we have to differentiate between pitch excursion and tonal scaling. The latter refers to a global level of register within the phrase while the former refers to the particular local implementation of an accent, for instance, from a low to a high turning point in Fo. Tonal scaling has received some attention in the past (Truckenbrodt, 2002, 2004; Fery and Truckenbrodt, 2005; Truckenbrodt and Fery, Ms; Fery and Kiigler, submitted) but is not dealt with in this chapter. The parameter 'pitch excursion' between two accentual targets has received relatively little attention with respect to intonational variation. The effect of broad and narrow focus on the alignment and pitch excursion of falling accents in Standard German has been investigated (Kiigler et al., 2003). It appears that initial pitch excursion is smaller in narrow focus since speakers tend to start higher. Tonal scaling remains the same in different conditions, but depends also on the position of the accent in the phrase. In his intonational comparison of urban German varieties Gilles (2005) suggests that pitch excursion functions as a dialect-specific cue, where the excursion of falling accents appears to be more reliable and less variant than that of rising accents. The absolute excursion size ranges from 6.43 semitones in Dresden to 9.56 semitones in Duisburg (Gilles, 2005, 165). In general, Gilles identifies dialect differences on the basis of excursion measurements, in particular, neighboring dialects have very similar excursions, and smaller excursions are found in Eastern German dialects while larger excursions are found in Western German dialects. To reduce the effect of speaker

139 specific pitch ranges, Gilles relates the individual measurements to the global pitch range of a speaker. In sum, the aspect of pitch excursion appears to be relevant to capture certain intonational variation between dialects.

6.2

Alignment of tones in Swabian and Upper Saxon German

6.2.1

Introduction

In this section, we will show that the phonetic implementation of pitch accents in terms of tonal alignment distinguishes phonological categories significantly from one another. On the basis of our phonological analysis of SwabG and SaxG intonational patterns in chapters 4 and 5, we will show that a given accent category aligns their tonal targets invariantly across contours, and that distinct accent categories differ significantly from one another. For SwabG we propose a tonal grammar that consists of three different nuclear contours and a modified one (cf. (30) in chapter 4). The main characteristic for that dialect is that each contour starts with a rising nuclear accent. Given the invariance hypothesis of tonal alignment we expect that the SwabG rising accent shows a similar and invariant tonal alignment behavior across nuclear contour type. Thus, the question to be answered is whether the alignment of tonal targets of rising accents in SwabG differ as a function of nuclear contour type. Indeed, we find a significant difference between rising pitch accents. The alignment measurements reveal two accent categories, a rising pitch accent (L*H) and a rising-falling one (L*HL), which is in line with our phonological analysis in chapter 4. For SaxG, on the other hand, we propose a tonal grammar that contains rising as well as falling accents in nuclear position. In addition, we assume a modification of base contours that result in L-prefixation, thus in a further phonological tonal category (cf. (27) in chapter 5). Given the invariance hypothesis, we expect that the three nuclear accents show a different pattern of tonal alignment. Thus, the question to be answered is whether the alignment of tonal targets in SaxG differ as a function of nuclear accent type. Our results show that the distinct pitch accents indeed differ significantly in their tonal alignment. In general, the analysis of alignment confirms our phonological analysis of SwabG and SaxG.

6.2.2

Analysis procedures

In our map task corpus (cf. chapter 2), we identified target words that contain a sonorant onset, and in closed syllables, at least one so nor ant coda consonant. According to German phonology, syllables with no coda contain a phonologically long vowel (e.g. Fery, 2000; Wiese, 1996), e.g. the second syllable in the word Arena [a.'Kei.na]

140 'arena'. Closed syllables consist either of a phonologically long or short vowel (ibid.), e.g. Mehl [ me:l] 'flour' and Lamm, ['lam] 'lamb' respectively. FQ has been extracted in Praat (Boersma and Weenink, 2005), using a Hanning window of 0.4 seconds length with a default 10 milliseconds analysis frame. Obvious errors of the Fo algorithm (e.g. octave jumps) have been corrected by hand, and the contour has been smoothed using the Praat smoothing algorithm (frequency band 10 Hz) to diminish microprosodic perturbations. All frequency measurements were done semi-automatically using a script that detects the lowest or highest Fo value within a given domain (for low tone detection, the domain is either the accented or post-accented syllable, for high tone detection, it is the accented word). The following segmentation points were identified in each intonation phrase (some of these points are identical to the ones in Atterer and Ladd, 2004): CO V0 VI Cl Ll Η L2 BT

the beginning of the onset of the accented syllable the beginning of the nucleus of the accented syllable the end of the nucleus (= the beginning of the coda if any) the end of the syllable rhyme (end of the accented syllable, i.e. after all coda consonants, if a word contains more than one) the first low tonal target (beginning of any rising part of an accent) the Fo maximum in rising or falling accents the second low tonal target (the end of any falling part of an accent) the boundary tone (relevant for excursion measurements)

5000

375a

2500-

1250 §

Ν I € CL

0.4

0.6 time [s]

Figure 6.1: Intonation phrase containing the target word Wohnwagen 'caravan' showing the segmentation and measuring points.

Figure 6.1 shows an example from the SwabG corpus with the above segmentation points, which is illustrated with the word Wohnwagen 'caravan'. The time of each of

141

these points as well as the frequency in Hz of the tonal labels have been collected, and are the basis for alignment and excursion calculations. Following Atterer and Ladd (2004), we calculated the absolute distance of tonal targets in relation to segmental boundaries. Tones (T) were calculated either with reference to V0 or Cl, cf. Equations 6.1 and 6.2. T

re/V0 = Τ - V0

(T = tone)

(6.1)

T

re/Cl

(T = tone)

(6.2)

= Τ - Cl

According to Equation 6.1, we calculated the alignment of accentual tones (L* and H*) in either dialect and of the leading low tone in SaxG (L+). According to Equation 6.2, we calculated the alignment of the trailing tones of either dialect (H and L). A negative value indicates that the tone (T) is realised before the segmental boundary, while a positive value indicates the realisation of a tone after the particular segmental boundary. It might also be possible to calculate the alignment relative to the beginning of the syllable CO. However, Atterer and Ladd (2004) suggest that the measurement according to Equation 6.1 fits more appropriately. We are aware of the fact that the exact way of expressing alignment data is debatable (in particular, see appendix A in Atterer and Ladd, 2004), yet we do not attempt to solve that issue here. In using the calculation according to Equation 6.1, we also provide data that is directly comparable to the one in Atterer and Ladd (2004) and other published data in, for example Ladd et al. (1999) and Arvaniti et al. (1998).

6.2.3

Results of tonal alignment in Swabian German

The alignment data of SwabG nuclear accents are reported in Table 6.1. The table shows, for each of the five nuclear contours proposed in chapter 4 separately, the mean alignment of the accentual low and trailing high tone. For complex accents (L*HL), the table shows also the mean alignment of the trailing L tone. The alignment of the accentual low tone is given relative to the beginning of the accented syllable's nucleus (V0), and that of the trailing tones (H and L) is given relative to the accented syllable's end (Cl). 2 The data of Table 6.1 are visualized in Figure 6.2. As can be seen, according to L alignment our claims of invariant alignment are proven, i.e. the initial low tone is aligned invariantly across contours. With respect to Η alignment, a rather variant pattern arises. The accentual low tone is realised near the beginning of the syllable nucleus. In four contours, the low tone is realised before the beginning of the syllable nucleus, while it is realised 1.4 ms on average into the vowel in case of the rise-fall contour (L*H L%). If we consider the alignment pattern of the accentual low relative to the syllable 2

We do not report standard deviations since the alignment measurement is a relational expression where standard deviations are inappropriate, cf. Atterer and Ladd (2004) for a similar approach.

142

Table 6.1: Mean alignment data for SwabG nuclear rising accents across contours. The columns show the distance in ms between an FQ label (L or H) and a segmental boundary (VO or CI). A negative value indicates that the Fo label occurs before the segmental label. η (137)

Contour L*H L% L*H L*H H% L*H+L L*HL H%

(rise-fall) (simple rise) (simple rise) (mod. rise-fall) (rise-fall-rise)

L*H(L%) /

/f^i'"

onset

nucleus

· > - '

H

refCl

-68.5 -10.4 10.7 -36.7 35.4

L

re/Cl -

184.3 224.5

\

\

\ \

\ L*H+L\ coda

9

*re/V0 1.4 -4.0 -5.5 -0.6 -24.1

L*H(H%) x·

X

•- y/

56 20 30 22

L

\

\

\

L*HL

postnuc. syllable

Figure 6.2: Schematic representation of SwabG accents with respect to alignment; no reference is given to pitch excursion.

nucleus in all contours together, we observe that the accentual low tone realisation ranges around the onset-nucleus boundary, with a total mean of -2.9 ms before the syllable nucleus. With regard to Η alignment, we observe a more varied pattern. In case of the L*H L%, L*H and L * H + L contours, the peak is on average aligned within the accented syllable, while it is realised in the postaccentual syllable in case of the L*H H% and L*HL H% contours. Further, the variation of mean Η-alignment appears to be rather large, from -10.4 ms relative to the accented syllables' end in the L*H contour to -68.5 ms in the L*H L% contour. The alignment of the trailing L tone in case of the complex L*HL / L * H + L pitch accents differ considerably between the two contours. While the trailing L is realised earlier relative to the accented syllable (184.3 ms after the accented syllable) in case of the modified contour L * H + L , it is realised later in case of the base contour L*HL H% (224.5 ms after the accented syllable). Given the invariance hypothesis of tonal alignment we would expect similar tonal alignment for similar accents across contours. Thus we test the question whether the alignment of rising accents in SwabG differs as a function of nuclear contour type. On the basis of independent samples t-tests we compare the alignment of each tone separately across contours. For the accentual low tone, we find no significant difference between contours, all comparisons are well above the 0.05 level (ρ > 0.05 for all L* across all contours). For the alignment of the trailing L tone between L * H + L and L*HL H%, no significant difference was found (df = 17, t = —1.43, ρ = 0.08), which is unexpected.

143

In case of peak alignment, the variable pattern that we find in the alignment data given in Table 6.1 is reflected by t-test comparisons between the contours. Peak alignment differs significantly between L*H L% and L*H (df = 46, t = —2.60, ρ < 0.006), between L*H L% and L*H H% (df = 55, t = -3.13, ρ < 0.001), between L*H L% and L*HL H% (df = 15, t = -3.88, ρ < 0.0007), and between L*H+L and L*HL H% (df = 15, t = —2.67, ρ < 0.008). In the other cases, no significant differences were found (p > 0.05). The analysis of tonal alignment reveals different results for the different types of tonal targets involved in the pitch accents. While the accentual low tones appear to align fairly invariantly, trailing Η alignment varies. According to our phonological analysis of chapter 4, we would predict that the rising pitch accent is identical in all SwabG intonation contours. The t-test comparisons between the contours in fact verified this prediction with respect to L alignment, since we do not find any significant difference in L alignment between the different contours. Interestingly, from a comparative point of view, the rather late alignment of the accentual low tone appears to be a German characteristic (cf. also the data in Atterer and Ladd, 2004, and our data on SaxG below). See also Truckenbrodt (2002) who briefly reports a similar alignment pattern of the low tone for German. Comparing our data with other languages, English, Dutch, Greek and Mexican Spanish align the low tone considerably earlier (Arvaniti et al., 1998; Ladd et al., 1999, 2000; Prieto et al., 1995). However, in line with Atterer and Ladd (2004), we observe variation of alignment of the L tone across German varieties. The particular alignment of L tones appears to be a dialect-specific cue. In fact, comparing our data with that of Atterer and Ladd (2004), we suggest a correlation between L alignment and geographical area, the nearer the syllable nucleus L is aligned, the more southern in origin the dialect (see section 6.3 below). The different alignment of the trailing L tone appears to point to the distinction that we have drawn between the base contour L*HL H% and the modified rise-fall L*H+L. Our proposal based on phonological assumptions is that the trailing L tone in cases of the modified contour is the result of a phonological process of L-suffixation (cf. chapter 4). On the contrary, the trailing L tone of a base contour belongs to the pitch accent itself. Prom that, we would expect different phonetic behavior of the two L tones. Though the difference in alignment of the trailing L is not significant between the two contours (p = 0.08), the direction of alignment difference, however, tends to point to the distinction between a base and a modified contour. That the trailing L alignment did not reach significance might be due to the very low number of tokens, in particular for the L*HL H% contour (n = 9). Possibly a larger group of tokens would show a clear difference. We can also interpret the non significant effect of trailing L tone alignment in a further direction. It could be the case that in case of the L*HL H% contour, trailing L alignment varies to such an extent that earlier aligned tokens might be interpreted as a modified rise-fall-rise, possibly a L*H+L H% contour, while later alignment represents the base contour, i.e. L*HL H%. If it turned out to be the case that we distinguish between different rise-fall-rise contours, it would strengthen our phonological analysis of SwabG contours in the assumption that a modification applies to all contours of the inventory. However, a different alignment in the rise-fall-rise contour is purely

144 speculative. Our dataset is just too small to show any such distinction, yet it would be a challenging task to investigate whether alignment differences of this kind result into two distinct intonation contours, and if so whether this would be perceived by the speakers of the dialect. If we consider the alignment of the trailing H, we observe a significant difference between the L*H+L and L*HL H% contours. The modified contour (L*H+L) aligns the trailing Η significantly earlier than the base contour (L*HL H%). Interestingly, there is no significant difference in Η alignment between the modified (L*H+L) and the base rise-fall (L*H L%) in which the the Η tone is realised well before the end of the accented syllable. 3 These two facts further support the assumption of a modification in case of the L*H+L, since the pitch accent involved is identical in structure to the base contour, and the complex trailing tone arises through modification. In sum, phonetically, the alignment data presents two independent arguments in favour of a phonological distinction between the accent shapes in the L*H+L and L*HL H% contours (though we only find a tendency and no significance for the trailing L alignment). A final remark has to be made on the different peak alignment behavior. According to our hypothesis, we do not expect significant differences in Η alignment between contours. For the simple rise (L*H H%), the particular late Η alignment (10.7 ms) might yet be due to a measuring paradox, since the peak is hard to determine if the contour is realised on a phrase-final trochee. This has been the case in 22 out of 30 instances in our corpus. Here, the pitch rises gradually until the end of the phrase, i.e. the trailing high target and the high boundary tone are identical. For the rise-fall-rise (L*HL H%), however, the only possible explanation for the difference in alignment appears to be that a different pitch accent (L*HL in contrast to L*H) is involved. Regarding the significant difference between the rise-fall (L*H L%) and the simple rise with high plateau (L*H), we have no explanation for the significantly later Η alignment in contrast to the rise-fall (L*H L%). In sum, the alignment data generally supports our phonological analysis of SwabG intonation pattern in chapter 4. We have presented phonetic evidence to support the view that the accentual low tone, thereby the first part of the nuclear pitch accent, is identical across contour type. Further, we have presented evidence that the assumption of a structural similarity between a rise-fall (L*H L%) and rise-fall-low-plateau contour (L*H+L) can be accounted for by assuming a phonological modification process of L-suffixation.

6.2.4

Results of tonal alignment in Upper Saxon German

The alignment data of SaxG nuclear accents are reported in Table 6.2. The table shows, for each of the six nuclear contours proposed in chapter 5 separately, the mean 3

Note that the alignment measurements relative to CI comprise syllables with phonologically long and short vowels. The calculation of mean duration of phonologically long vs. short vowels reveals a difference in vowel length, ä?v: = 161.5 ms, η = 105 vs. xv = 91.0 ms, η = 90.

145

alignment of the L-prefix (L+) and the accentual tone (T*, either H* or L*) relative to the syllable nucleus (VO), and of the trailing tone (T, either L or H) relative to the end of the accented syllable (Cl).

Table 6.2: Mean alignment data for SaxG nuclear accents across contours. The columns show the distance in ms between an Fo label (L+, T* and trailing T) and a segmental boundary (VO or Cl). A negative value indicates that the Fo label occurs before the segmental label. Contour H*L H*LH% L+H*L L+H*LH% L*H L*H H%

(simple fall) (fall-rise) (mod. simple fall) (mod. fall-rise) (simple rise) (simple rise)

η (133) 42 3 37 10 15 26

L

+refV0 -

-

-14.3 -47.0 -

-

T

*refY 0 15.7 14.4 119.9 110.9 3.0 0.6

^Ve fCl 65.9 -25.4 144.0 182.3 14.5 10.7

As can be seen, each of the three nuclear accents involved (H*L, L+H*L and L*H) appear to be clearly separated from each other, which confirms our hypothesis of invariant alignment of different tones within a language, cf. the schematic illustration in Figure 6.3. In all accents, the accentual tone (T*) is realised on the accented vowel, the trailing tone in the postnuclear syllable (except for H*L H%, which may be due to the following high boundary tone). "v.

/ -

. . -'L*H

y -

/

/

s

onset

H*L nucleus

coda

~~ C+H*L

postnuc. syllable

Figure 6.3: Schematic representation of SaxG accents with respect to alignment; no reference is given to pitch excursion. If we compare the falling accent (H*L) with the corresponding modified one (L+H*L), we observe that the accentual high tone differs in alignment; it is realised about 15 ms after the beginning of the vowel in case of the falling accent, and about 100 ms later than in case of the modified accent. Accordingly, the trailing low tone is aligned considerably later in the modified accent than in the base accent. If we compare the rising accent (L*H) with the modified falling one (L+H*L), we observe that the initial low tone differs in alignment; the accentual low (L*) is realised at the beginning of the vowel, the leading low tone (L+) is realised on the onset of the syllable (-14.3 to -47.0 ms). Accordingly, the trailing high tone in case of the rising accent is realised later than the accentual high tone of the modified falling accent. In order to test whether the individual accents involved in different contours are identical (e.g. H*L in a simple fall and rise-fall contour), we conducted independent

146 samples t-tests. We find no significant difference for within accent comparisons for none of the three pitch accents (ρ > 0.05 for L+ r e ^yo> T* r e ^yQ and T^refQi alignment). This allows us to group the accent data into three groups, i.e. a falling accent (H*L), a rising accent (L*H) and a modified falling one (L+H*L). In order to test the invariance hypothesis that the accents differ in alignment across accent type, again, we conducted independent samples t-tests, yet for the grouped data. The alignment of the accentual high tone differs significantly between the falling (H*L) and modified falling (L+H*L) accent (df = 90, t = - 8 . 9 3 , ρ < 0.0001). The alignment of the accentual/leading low tone differs significantly between the rising (L*H) and the modified falling (L+H*L) accent (df = 86, t = -2.21, ρ < 0.014). Finally, the alignment for the trailing low tone between the falling (H*L) and modified falling accent (L+H*L) differs significantly ( L r e ^ c i alignment between H*L and L+H*L df = 85, t = -3.31, ρ < 0.0006). We have conducted alignment measurements in order to provide phonetic evidence supporting our phonological analysis of SaxG intonation patterns in chapter 5 where we proposed three distinct nuclear pitch accents, H*L, L*H and L+H*L. The last, L+H*L, we claim, is a result of a phonological process of L-prefixation (cf. Gussenhoven, 2004). We have argued in chapter 5 that the modified pitch accent embodies properties of both basic pitch accents, i.e. the rising part ( L + ) shares considerable properties of the simple rise, while the falling pitch accent (H*L) is structurally similar to the simple fall. Given the invariance hypothesis, we expect that the three nuclear pitch accents of SaxG differ fundamentally in their alignment behavior, which is proven by our analysis of tonal alignment. Moreover, we have shown that the individual pitch accents maintain their alignment pattern no matter which contour an accent is realised in. For instance, a simple rising accent (L*H) aligns its two tones similarly in a L*H and L*H H% contour, we did not find any significant differences for a pitch accent with occurrences in different contours. The result corroborating our phonological analysis of chapter 5 is that the L+H*L accent differs significantly both from a simple falling accent with respect to its accentual high and trailing low tone and from a simple rising accent with respect to the leading low and accentual high tone. We have shown for SwabG that the accentual low tone does not differ significantly across nuclear contours. If the leading low tone of the modified rise-fall accent would be a L*, we would expect no significant difference in L alignment with the simple rise (L*H), as evidenced by our SwabG findings. The leading L tone is aligned significantly earlier (i.e. in the onset) than the accentual L* (i.e. at the beginning of the nucleus) (cf. Table 6.2), it is thus different from the accentual low tone. At the same time, we expect a significant difference with respect to Η alignment when we compare the modified contour with its base contour, which we indeed observe (cf. Table 6.2). In case of the simple fall, the accentual Η aligns around 15 ms after the beginning of the vowel. In case of the modified simple fall, the Η tone is aligned around 110 to 119 ms after the beginning of the vowel. Given an average nucleus duration of 158 ms for phonologically long vowels,4 the Η tone aligns well within the 4

The calculation of mean duration of phonologically long vs. short vowels reveals a difference in vowel length, xv: = 158.1 ms, η = 91 vs. x\ = 93.4 ms, η = 109.

147 accented vowel, it is thus an accentual high target. This property is the structural similarity to the simple fall. Regarding the rise-fall accent and its marked status in SaxG intonation, we have shown that this accent differs with respect to both its rising part and its falling part. The significant difference is between the rising and the falling accent independent of each other. This result clearly supports our analysis of the rise-fall accent that it is its own phonological category.

6.3

A comparison of accent realisation

6.3.1

Introduction

The final part of this chapter is about an attempt to model dialect variation in intonation. We will show that phonological identical contours differ with respect to phonetic implementation across language varieties; similarities can also be identified. We will also show that the accent shape on the accented syllable of functionally identical, yet phonologically distinct contours contribute to the dialectal characteristics of accentuation. In the introduction of this chapter, we have pointed to two levels on which accent implementation may vary across languages, i.e. tonal alignment and pitch excursion. With respect to tonal alignment, we have shown in the preceding sections that within a given language variety, distinct pitch accents can be distinguished on the basis of their tonal alignment patterns. Atterer and Ladd (2004) showed that the individual alignment patterns are dialect-specific and thus predictable on the basis of quantitative phonetic implementation rules. With respect to pitch excursion, Gilles (2005) showed that German dialects differ with regard to the base line from which speakers tend to start, and with regard to the pitch range speakers use. These facts point to the relevance of pitch excursion, or pitch range to be a dialect-specific cue as well. Given the findings of Gilles (2005) we expect significant differences in pitch excursion size between the two dialects. Our model of pitch accent realisation takes into account two factors, tonal alignment and pitch excursion. The model is based on a bottom-up approach to model variation in accent realisation. Our data is fairly uncontrolled, yet we will show that the two factors discussed above play an important role in quantifying pitch accent realisation. We assume that if the model were tested in a controlled setting that we could certainly improve the significance of the two factors as well as rule out disturbing influences of uncontrolled material. In the next section, we illustrate our measurement and calculation procedure of accent realisation. Thereafter, we present the results of comparing different contours with each other.

148

6.3.2

Analysis procedures

The analyses and measurements reported here are based on the data analysed above, section 6.2.2. The only contour that is tonally equivalent between SwabG and SaxG is the simple rise. Yet, only the simple rise with a high plateau (L*H) is comparable since in the other case (L*H H%), we cannot reliably identify the high trailing tone (cf. above). The comparison concentrates therefore on the simple rise contour to illustrate the differences between SwabG and SaxG in phonetic implementation. Alignment: To express the alignment of tones in a comparable way, we need to express the actual alignment measurements presented in the previous section in terms of the relative distance between the tonal target and a segment or syllable. Based on the reported findings, low tone alignment is best analysed with reference to the beginning of the accented vowel (VO). The relative low tone alignment has yet to be related to the accented syllable in order to compare the data. Therefore, we choose to calculate L alignment by means of the distance between the low tone and the end of the onset consonant (VO) divided by syllable duration (Cl — CO) (cf. Equation 6.3a). Alignment of the high tone is best accounted for by calculating the relative distance between the high tone and the end of the accented syllable (Cl) divided by the duration of the accented syllable (Cl — CO). The end of the accented syllable appears to be a reasonable segment boundary, since the high tone may either be realised in the accented syllable, thus before the end, or in the next syllable, thus after the end (see also Atterer and Ladd, 2004, for this measurement). See Equation 6.3b.

(a) L(%) =

Xsyll

* 100

(b) H(%) =

+*"» 0.05). Therefore, mean onset duration is based on mean onset duration of SwabG and SaxG onsets (Xonset = 76.9 ms, η = 220). An independent samples t-test for syllable duration reveals a significant difference between SwabG and SaxG accented syllables (df — 206, t = 1.79, ρ < 0.05). Therefore, tonal alignment with respect to syllable duration is based on each dialect's mean syllable duration. The means of SwabG alignment data are XL = —2.9 for L-alignment, and XH = — 10.4 for Η alignment (cf. Table 6.1 above). The data for L are grouped, since we did not find any significant difference for L-alignment across contour type. The means of SaxG L*H are xL = 3.0 and xH = 14.5 (cf. Table 6.2 above). Excursion: For pitch excursion, we first have converted the frequency measured in Hz into semitones. We follow Nolan (2003), who has convincingly demonstrated that the semitone scale fits best the intonational equivalence (note also Ladd, 1996, 260ff for the notation of semitones with respect to pitch range). Also, the semitone

149 scale allows for a normalization across gender. In terms of Hertz, a female voice has a higher pitch level than a male voice. This is normalized in the semitone scale since this scale is based on frequency intervals, taking into account that a frequency interval of 50 Hz between 50 and 100 Hz is not equal to an interval between 200 and 250 Hz (see Reetz, 1999, for instance, for technical details of the semitone scale). The conversion into semitones is made according to Equation 6.4 with an arbitrarily chosen reference of 100 Hz.

=

(6 4)

·

Based on semitones, we calculated the pitch excursion by subtracting the low pitch of a pitch accent from the high pitch value (Equation 6.5a) irrespective of accent type. We concentrated on the rise of rising accents. Further, we calculated the mean low tone in relation to the mean low boundary ( B T ) (Equation 6.5b). This was done in order to provide a baseline for the data plot. (α) Ε [st] = H s t - L s t

6.3.3

(b) L [st] = L s t - BT,st

(6.5)

Results

Table 6.3 presents the results of the alignment and excursion calculations. An independent samples t-test for Excursion (E) reveals a significant difference between SwabG and SaxG (E [st] df = 21, t = - 2 . 4 4 , ρ < 0.05); for the baseline (L tone in relation to the low boundary tone), no significant difference between SwabG and SaxG can be found (L [st] ρ > 0.05). For alignment, we find a significant effect for Η alignment (df = 71, t = —1.81, ρ < 0.05), yet no effect for L* alignment (p > 0.05). Table 6.3: Data of L and Η alignment in percent of syllable duration, and L minimum and Excursion (E) in semitones.

SwabG SaxG

6.3.4

L*H L*H

L [st] 2.45 2.25

L [%] 24.4 28.7

Ε [st] 7.01 8.55

Η [% 92.0 104.4

Realisational differences of pitch accents - a model

Figure 6.4 illustrates the difference between the SwabG and SaxG simple rise accents with respect to tonal alignment and pitch excursion (based on the data of Table 6.3 above). As can be seen, excursion size differs between the dialects. The lower range is fairly similar between the dialects, SwabG speakers implement a higher rise than SaxG speakers.

150

Figure 6.4: A comparison of alignment and excursion between SwabG and SaxG simple rising accents in the simple rise contour (L*H).

Figure 6.4 illustrates that the dialects differ in some of the parameters. In terms of alignment, the differences are in the realisation of the high tone. And in terms of pitch, the dialects differ in excursion size. In addition to dialect-and language-specific alignment of tones (Atterer and Ladd, 2004), the parameter 'pitch excursion' appears to signal dialect-specific intonational features as well. For L-alignment, we find no significant difference between SwabG L* and SaxG L* (p > 0.05), meaning that the position of the accentual low tone relative to the onset and syllable duration is identical in the two dialects. Comparing this result to Atterer and Ladd (2004) who have found a significant difference between Northern and Southern (Bavarian) German, we can conclude that for rising accents occurring in a L*H nuclear contour, there is no difference between South Western and Middle Eastern German. Recall that Atterer and Ladd (2004) analyse prenuclear rising accents. The similar implementation of accentual low tones might be an indicator of a perceived similarity between SwabG and SaxG (many of our informants have reported that these dialects do not differ that much intonationally). Our model comprises two factors of pitch accent realisation that encounters the relation of pitch and alignment in and around the accented syllable. Gilles (2005) proposed a different model expressing these factors. Gilles expresses tonal alignment in terms of distance relations of the tone in relation to the beginning and end of the accented syllable, thus not in relation to onset nucleus boundary which has been proposed by Atterer and Ladd (2004), and which we also assume. Gilles further calculates the accentual Fo minima and maxima in relation to global pitch range, where global pitch range is based on calculations of 40 randomly chosen, chronologically following intonation phrases of his speech corpus. Given the observation that one might identify a speaker's dialect on the basis of a single utterance, we doubt a calculation procedure that relies on global means of pitch range. We believe that our proposed calculation method reveals obvious differences between the dialects, though we are aware of the fact that different measurements could show even better results, e.g. different normalization calculations or alignment calculations. We also believe that a calculation which is based on the accented syllable reflects the aspects of accent realisation more properly since it is the syllable which is the domain of tonal implementation - at least for intonation languages.

151 There might be a further objection against global pitch range, that is, the question of defining the relevant measuring points. Gilles (2005, 85) does not explicitly discuss the choice of measuring points, i.e. which highest and lowest points to measure. Möhler and Mayer (1999) propose a calculation of pitch range based on pitch accents and boundary tones. A different proposal made by Patterson (2000) excludes the first and last pitch accent for the calculation of pitch range. Given these obvious methodological differences, global pitch range and its relation to pitch accent implementation remains an open issue. If we apply this model to illustrate the FQ contour on the accented syllable for functionally identical, yet phonologically distinct, contours we identify crucial differences in the phonetics of accented syllables across dialects. Given that the accented syllable is the perceptually salient and most prominent part of the nuclear contour, the differences in phonetics reveal the relevant cues on which speakers base their perception on. Consider for instance the modified SwabG rise-fall (L*H+L) and the modified SaxG simple fall (L-|-H*L). Functionally, these contours share their expression of narrow focus (cf. chapters 4 and 5), and with respect to intonational meaning, these contours convey a meaning of 'significant ADDITION' (cf. chapter 7 ) . The shape of the respective accent is illustrated in Figure 6.5, where differences in tonal alignment and excursion sizes become obvious.

SwabG L*H+L SaxG L+H*L

20

40

60

80

100

120

relative syllable duration ( r r}

Figure 6.5: A comparison of alignment and excursion between rise-fall contours: SwabG L*H+L, SaxG L+H*L.

The FQ shapes on the accented syllable shown in Figure 6.5 are similar, yet not identical. Despite the difference in phonological form, there are subtle differences in the phonetic realisation on the accented syllable regarding the alignment of the L and Η tone and pitch excursion. The SaxG accent shape is realised earlier than the SwabG one. In addition, the SaxG accent shape has a larger excursion than the SwabG one. An even more distinct picture arises when we are comparing the case of neutral accentuation between SwabG and SaxG; for SwabG, we have analysed a rise-fall (L*H L%), and for SaxG a simple fall (H*L). The difference of accent shape on the accented syllable is shown in Figure 6.6.

152

relative syllable duration

| ]

Figure 6.6: A comparison of alignment and excursion in neutral accentuation: SwabG L*H L%, SaxG H*L.

For SwabG, the accent shape of the accented syllable is rising in neutral accentuation, for SaxG it is falling. This is apparent from our phonological analysis, where we analyse a rising pitch accent for SwabG (L*H) and a falling one for SaxG (H*L). Thus, one can argue that differences in phonological form predict differences in phonetic form. This objection does not hold, however, if we compare the data in Figure 6.6 with neutral accentuation in Standard German. According to Grabe (1998a) Standard German in neutral accentuation exhibits a simple falling accent (H*L) (cf. also Fery, 1993), which is phonologically similar to the SaxG simple fall. The phonetic form, however, differs from SaxG since Grabe showed that falling accents align their accentual high tone at the right edge of the syllable rhyme. In Figure 6.7, we have plotted Grabe's alignment data (without reference to excursion since no data are available for that) against the SwabG and SaxG comparison of Figure 6.6.

Figure 6.7: A comparison of alignment and excursion in neutral accentuation between SwabG L*H L%, SaxG H*L and Standard German H*L.

As can be seen in Figure 6.7, the Standard German simple fall is fairly similar in phonetic form on the accented syllable to the SwabG rise-fall; note that these two accents differ in phonological form! A phonologically identical accent, on the contrary, appears in a different accent shape on the accented syllable, cf. the SaxG simple fall compared to the Standard German one. Thus, the phonetics of the accent shape on

153 the accented syllable is a crucial property of a given language variety irrespective of phonological form of the contour. An interesting question that we leave for future research is the perceptual relevance of these aspects of accent realisation. One might investigate how much the accent shape of the accented syllable influences the perception of a phonological tone. The acoustic measurements at least appear to point to specific dialect features as regards the implementation of accents on the accented syllable.

6.4

Summary

This chapter is concerned with phonetic aspects of tonal implementation, in particular with tonal alignment and pitch excursion. For both dialects, we show that with respect to tonal alignment the proposed phonological tones differ significantly from one another. In our tonal analysis we propose five distinct intonation contours in SwabG, which all start with a rising pitch accent L*H. We show that the alignment of the accentual low tone does not differ significantly across contour types. On the contrary, in our tonal analysis of SaxG we propose six distinct nuclear contours with differing nuclear accents. According to the invariance hypothesis of tonal alignment, we show that each nuclear accent differs significantly from one another. In a comparison between the dialects, we identify two parameters that, taken together, allow for quantifying intonational variation with respect to tonal implementation across language varieties. Our model combines the aspect of tonal alignment with respect to the accented syllable and the aspect of pitch excursion for implementing pitch accents.

A note on intonational meaning

7

7.1

Overview

This chapter attempts to analyse the meaning conveyed by the individual nuclear contours proposed for SwabG and SaxG. We provide an analysis of intonational meaning following the contour based model proposed by Gussenhoven (1984). Our analysis is far from exhaustive and we do not pretend our analysis to be comprehensive. We do believe, however, that we show that Gussenhoven's model of intonational meaning is extendable to SwabG and SaxG. Further, we believe that our data and analysis provide two arguments that speak against a compositional approach to intonational meaning (Pierrehumbert and Hirschberg, 1990). A comparison between SwabG and SaxG nuclear contours with respect to intonational meaning reveals striking similarities despite the obvious tonal differences between the two dialects. The first part of this chapter (section 7.2) introduces briefly the theory of intonational meaning, first the contour approach put forward by Gussenhoven (1984) and second the compositional approach by Pierrehumbert and Hirschberg (1990). In section 7.3, we present our analysis of intonational meaning separately for SwabG (section 7.3.1) and SaxG (section 7.3.2). The chapter closes with a discussion in section 7.4 in which we argue that Gussenhoven's model more appropriately captures the intonational characteristics of SwabG and SaxG than the compositional approach of Pierrehumbert and Hirschberg (1990).

7.2

Intonational meaning

The primary goal of the study of intonation is to define "which intonation patterns are linguistically distinct and which count as variants of the same pattern" (Pierrehumbert, 1980, 31). According to Pierrehumbert, one might approach this primary goal from two different angles: "One approach attacks the problem by attempting to deduce a system of phonological representation for intonation from observed features of F 0 contours. After constructing such a system, the next step is to compare the usage of F 0 patterns which are phonologically distinct. The contrasting approach is to begin by identifying intonation patterns which seem to convey the same or different nuances. The second step is to construct a phonology which gives the same underlying representation to contours with the same meaning, and different representations to contours with different meanings." (Pierrehumbert, 1980, 31)

In the main part of this study, we have concentrated on the first approach, that is, on the basis of our speech data we have argued for a number of individual contours to represent the phonological inventory of SwabG and SaxG intonation.

155

In this chapter, we will turn to the second approach, that is, to define the meanings that the individual contours convey, and thereby provide evidence in favour of our phonological analysis. We are aware that this cannot be more than a tentative enterprise, yet we believe we provide a fruitful basis for future research in this direction. In the next two sections, we introduce two different theories of intonational meaning. We will follow the contour approach (Gussenhoven, 1984). A theory of intonational meaning is anchored in the assumption that intonation contributes to the interpretation of individual utterances and utterances in discourse. Any utterance conveys a kind of meaning that a speaker intends to express in a particular situation or part of a discourse to keep the discourse running. A speaker and a hearer share a certain amount of knowledge about the world, themselves and the particular discourse. This is called background (Gussenhoven, 1984) or mutual beliefs (Pierrehumbert and Hirschberg, 1990), and the choice of a particular intonational tune of a speaker is always related to this shared knowledge. The crucial distinction between the contour-based and the compositional approach of intonational meaning is in the interpretation of the intonational tune in relation to the background.

7.2.1

A contour based approach

In chapter 3 , section 3 . 2 . 5 , we have outlined the basic assumptions of Gussenhoven's theory of intonational meaning. Recall from that overview that Gussenhoven ( 1 9 8 4 ) assumes that a speaker basically has three options to refer to the background, i.e. ADDITION, S E L E C T I O N and R E L E V A N C E T E S T I N G . This reference to the background represents the basic meaning a speaker may express. Gussenhoven further assumes that each individual meaning is expressed by means of a distinct nuclear contour, i.e. a nuclear contour is thus viewed as a morpheme. For Standard British English, Gussenhoven ( 2 0 0 4 ) proposes that the fall, the fallrise, and the rise are the basic contours that relates to the meaning of A D D I T I O N , S E L E C T I O N and R E L E V A N C E T E S T I N G , respectively. A base contour might be extended by modifications like D E L A Y , or L-prefixation (Gussenhoven, 2 0 0 4 , 3 0 7 ) . The assumption of a distinction between base contours and modified contours is very powerful in that it explains why a bundle of contours share identical tonal parts. For instance, the modification L-prefixation shifts the original nuclear contour to the right and inserts a new tonal target, a L tone. This occurs systematically to all three base contours, where each modified contour maintains the structural characteristics of the base contour. Gussenhoven assumes three modifications, L-prefixation, Stylisation and Half-Completion. We concentrate our discussion here on L-prefixation, since this modification accompanied with the meaning proposed for English appears to occur in our data as well, as we will show below. According to Gussenhoven, L-prefixation adds a nuance of 'significance' to the meaning of the base contour, it expresses the meaning of "very non-routine, very significant" (Gussenhoven, 1984, 218). Consider for instance a simple fall in British English, that conveys a meaning of ADDITION, i.e. a speaker adds a particular information to the background of the speaker and hearer. In case of the L-prefixed simple fall, a speaker puts particular emphasis on the information to be added to the back-

156 ground, which could be paraphrased as "I hereby tell you that it is this particular information that is to be put in the background". In his 2004 book, Gussenhoven refers briefly to Dainora (2002) motivating the contour approach to intonational meaning. Dainora conducted a corpus-based study calculating the probability of tone sequences in nuclear contours. Her main result is that the nuclear accent appears to be "a strong predictor of boundary tone" (Dainora, 2002, 235), which means that for every pitch accent there is one particular boundary tone that the pitch accent is most likely combined with. According to Dainora's analysis, each pitch accent might be combined with other boundary tones as well, yet with less probability. On the basis of probability, Dainora concludes therefore that the nuclear accent in combination with a boundary tone forms a unit. However, from our point of view, a probability account cannot motivate this unit since less probable contours do exist - as is also shown in Dainora's data. Gussenhoven himself refers to contours that occur less frequently, yet they have to be accounted for and are to be included in the tonal grammar of English. Still, from our point of view, the contour approach of Gussenhoven (1984, 2004) appears superior to the compositional model since it very nicely fits the variability of intonational contours in a relatively simple meaningful analysis system. In addition, our proposed analysis of SwabG and SaxG appears to be captured by means of the contour-based approach, which we will show below.

7.2.2

A compositional approach

Pierrehumbert and Hirschberg propose a theory of intonational meaning that is based on the communicative function of intonation that "a basic goal of speaker S is to modify what (S believes) a hearer Η believes to be mutually believed" (1990, 286). Thus intonation functions to highlight information (in the sense of Bolinger, Ladd and others), yet the basic meaning of intonation refers to how a speaker uses the intonational function of highlighting information with the goal of modifying the mutual beliefs in a speaker-hearer communicative setting. A certain tune that a speaker is choosing conveys a particular relationship between an utterance and the current mutual beliefs of the participants of a conversation. According to Pierrehumbert and Hirschberg, these relationships are compositionally interpretable from the independent meaning of pitch accents, phrase tones and boundary tones. These categories take scope over different domains, i.e. pitch accents take scope over the lexical material that is accented, phrase tones, over intermediate phrases, and boundary tones, over the whole intonation phrase. Taken together, "these intonational features can convey how S [the speaker, F.K.] intends that Η [the hearer, F.K.] interpret an intonational phrase with respect to (1) what Η already believes to be mutually believed and (2) what S intends to make mutually believed as a result of subsequent utterances" (Pierrehumbert and Hirschberg, 1990, 288). Table 7.1 summarizes the individual meanings that Pierrehumbert and Hirschberg ascribed to the tonal events (for a thorough review of intonation and meaning see Mayer, 1997). For instance, a high pitch accent signals that the accented information is new with respect to the mutual beliefs. In combination with a low phrase and

157 boundary tone (Η* L L%), the meaning is to add the new information to the "mutual belief space" (Pierrehumbert and Hirschberg, 1990, 290). If, on the other hand, a high pitch accent is followed by a high phrase and boundary tone (Η* Η H%), Η H% "'question' the relevance of that information" (Pierrehumbert and Hirschberg, 1990, 290) that H* adds to the mutual beliefs. If a high pitch accent is followed by a high phrase but low boundary tone (Η* Η L%), again, the information highlighted by accentuation is to be added to the mutual beliefs, but the phrase has to be interpreted with reference to other sentences in the discourse. Table 7.1: Summary of intonational meanings for pitch accents, phrase tones and boundary tones in American English according to Pierrehumbert and Hirschberg (1990). Tonal item Meaning Pitch accents H* new in the discourse (p.289) L* salient without predication (p.292) L+H identifying a relative scale (p.302) L*+H lack of prediction and evoke a scale (p.296) L+H* accented item should be mutually believed (correction/contrast) (p.296) H+L identifying an inference path (p.302) H*+L open a reference path between new (accented item) and old information (mutual beliefs) (p.298) H+L* instantiation of the open expression is already present among H's mutual beliefs (p.300) Phrase tones H— forming a larger composite interpretative unit with the following phrase (p.302) L— separation of the current phrase from a subsequent phrase (p.302) Boundary tones H% forward reference; hierarchical relationship between intentions underlying the current utterance and subsequent one(s) (p.305) L% interpret the utterance without any reference to subsequent utterances (p.305) The interesting aspect of this theory is its compositionality, since it takes seriously the theoretical assumptions of Pierrehumbert's tone-sequence-model into a theory of intonational meaning; the assumption is that the individual tones are morphemes. However, the approach of Pierrehumbert and Hirschberg has some drawbacks which, to some extend, have been revised by Hobbs (1990). Hobbs simplified the Pierrehumbert and Hirschberg proposal grouping the different tonal events: accent tones (the

158

starred tones) and leading tones on the one hand, and trailing tones, phrase tones and boundary tones on the other hand. The meaning attributed to the respective group can be classified as in Table 7.2. Based on this tonal grouping Hobbs achieves a more coherent interpretation of the meaning attributed to the tones. Table 7.2: The meaning of intonation according to Hobbs (1990), adapted from Mayer (1997, 53). Meaning of I (accent + leading tones) II (trailing, phrase, + boundary tones)

Η salient + new incomplete, open-ended

L salient + not new underspecified w r t completeness

However, the approaches of Pierrehumbert and Hirschberg (1990) as well as Hobbs (1990) do not appear to be fully spelled out. This concerns particularly the interpretation of certain contours with respect to surrounding discourse, and the interpretation of the phrase tone in general. For instance, Pierrehumbert and Hirschberg (1990) remain somewhat vague in the interpretation of the Η* Η L% contour, cf. (1). The authors contrast this contour with H* L L% and Η* Η H%, and argue for intonational compositionality, since in all three contours the H* accent signals an attempt to modify the mutual beliefs of the hearer (1990, 291). To our point of view, however, the interpretation of the Η* Η L% contour remains unclear. (1)

Η* Η L% contour in Pierrehumbert and Hirschberg (1990, 291) Wally: Mostly they just sat around and knocked stuff. You know. The school Η* Η L% Other people Η* Η L%

This does not become much clearer when Pierrehumbert and Hirschberg discuss the meaning of phrase and boundary tones. In that section, the authors do not refer to intonation phrases like (1) but to lists or continuations such as (2). Here, the phrase tone takes scope over a whole intermediate phrase. The meaning of (1), however, is not explained with examples like (2). What makes the vague point even weaker is that the examples given for the Η* Η L% contour (cf. (1)) are particularily ill-chosen since they cannot show the tonal implementation due to an insufficient number of syllables. On phrase-final or phrase-penultima syllables the phenomenon of a high plateau cannot convincingly be shown. (2)

Phrase accent in listings (Pierrehumbert and Hirschberg, 1990, 302) Do you want apple juice or orange juice Η* Η H* L L%

159

The authors themselves point to the weakness in interpretation: "However, [...] we are not able to present such a decomposition yet" (Pierrehumbert and Hirschberg, 1990, 301) and "we have presented the beginning of a compositional theory of meaning of intonational contours" (Pierrehumbert and Hirschberg, 1990, 308, emphasis' ours). In particular, this appears to concern the phrase as well as boundary tones and the scope that the individual tones take. The second main objection against a compositional approach is that, to our point of view, a holistic or contour based approach has not convincingly been argued against. In particular, Pierrehumbert and Hirschberg (1990, 293) argue that examples like (3) have allowed Liberman and Sag (1974) to argue for a holistic contradiction contour. (3)

A: Let's order the Chateaubriand for two. Β: I don't eat beef. L* L* L H% (Pierrehumbert and Hirschberg, 1990, 293)

Compare (3) with an example of the contradiction contour of Liberman and Sag (1974) in (4). Note that the contradiction contour has an "initial rise and fall, uncorrelated with word stress" (Liberman and Sag, 1974, 421), wile the contour in (3) has a prenuclear low accent. Thus, reading Liberman and Sag (1974) carefully shows that the proposed contradiction contour in (4) has nothing to do with Pierrehumbert and Hirschberg's example in (3). Thus, it is our opinion that the argument against holistic contours conveying intonational meaning fails. Our analysis below will show that a compositional approach also fails to cover the meanings of SwabG and SaxG contours. (4)

Elephantiasis isn't incurable !

(Liberman and Sag, 1974, 420)

7.3

A n a t t e m p t of analysing intonational m e a n i n g

7.3.1

The meaning of Swabian German nuclear contours

In this section, we will analyse the meaning of SwabG nuclear contours according to the contour based approach (Gussenhoven, 1984). According to our analysis in chapter 4, in SwabG, we distinguish between three base contours and a modified one. For each contour separately, we will analyse the meaning that the contour conveys. The differences in meaning in SwabG nuclear contours arise in the postnuclear part, since all nuclear contours start with an accentual low tone. As we have mentioned in the introduction of this chapter, our analysis is far from exhaustive. The analysis provided here attempts only to illustrate our proposal how different meanings might be correlated with which contours. We propose that the rise-fall conveys a meaning of A D D I T I O N , the rise-fall-rise SELECTION and the simple rise R E L E V A N C E T E S T I N G .

160 For the rise-fall (L*H L%), we propose that it conveys a meaning of ADDITION. TO illustrate this, consider the passage (5). Both speakers are at the very beginning of the second map task (cf. appendix A). The instruction giver (speaker B) starts with explaining where the starting point for this particular map is, and outlines the route on the map towards the first symbol Mehl 'flour'. In line 10, Β finishes the first part of instruction with a declarative sentence that the first symbol MEHLsack 'sack of flour' occurs at that particular position on the map. 1 B: 2 3 A: 4 5 B: 6 7 A: 8 B: 9 A: 10 B:

also der start unsrer karte befindet sich links Oben so the start of our map is top left ogfähr anderthalb Zentimeter (-) vom RAND we [g about one and a half centimetres from the margins [m=hm uhm den h[ab ich ja I do have this yes [nach rechts und (.) nach unten to the right and downward isch des bei dir genau=s GLEIche do you have exactly the same m=hm uhm so die strecke beginnt senkrecht nach UNten so the route is starting vertical down sen[krecht (-) nach links vertical to the left [da isch (.) rechts davon so η MEHLsack L*H L'/. on the right of it there is like a sack of flour

From the point of view of information structure, the phrase in line 10 provides new information, a case of broad focus. The instruction giver might assume that the instruction receiver also has a symbol 'sack of flour', yet because the speakers perform the second map task, they already know that the maps differ. Thus, sentences as in (5), line 10, convey a meaning that we could paraphrase as "I hereby provide (new) information which I want you to add to our background". This meaning is what Gussenhoven (1984) has analysed as ADDITION in British English. The information provided by the speaker is intended to update the background of the speaker and hearer. Given that the sentence in (5), line 10, is realised with the rise-fall contour, we conclude that the rise-fall in SwabG conveys the meaning of ADDITION. For the rise-fall-rise (L*HL H%), we propose a meaning of SELECTION. To illustrate this, consider the passage (6). The subjects have received their instructions about the map task and are about to start. The situational background of the passage in (6) can be characterized as follows: the subjects know that they both have a map with a number of individual symbols drawn on it; speaker A functions as the instruction giver, speaker Β as the instruction receiver, whose task it is to draw the route on the map. We assume that the instruction giver assumes that the two maps are equal (a fact that the speakers should not have any doubts about at the beginning of the

161

task). The instruction giver intends to 'set the scene' for the start of the map task (cf. line 3). (6)

1

A:

2

B:

3 4

A:

—> 5

6

Β:

jetzt kann=ich loslegen now I can start also gut (.) ok dann [fang=ma an let's start [ok (-) ok s=schatz mir befinde uns jetz=hier an dem NONenweiher L*HL H'/. darling we are here at the idyllic pond m=he uh=m

With the target sentence in line 5, speaker A wants to remind the instruction receiver where to start. Given the assumption of the instruction giver that the maps are identical, the start position around the symbol 'idyllic pond' is in the situational background of the speaker and hearer. Speaker A selects this information from the background, and reminds speaker Β of this by setting the scene for the start. The sentence in line 5 might be paraphrased as "I remind you of this fact" (cf. Gussenhoven, 1984, 204). This meaning is what Gussenhoven (1984) has analysed as SELECTION. Given that the sentence in (6), line 5, is realised with the rise-fall-rise contour, we conclude that the rise-fall-rise in SwabG conveys the meaning of SELECTION. SELECTION means, that a certain element or information is selected from the speaker-hearer-mutually-shared conversational background. The difference between SELECTION and ADDITION is in direction: while for ADDITION, an element or an information is put into the background, for SELECTION, it is taken from the background. Generally, the contour type that is associated with SELECTION conveys a questionlike meaning in the sense that a particular kind of background is discussed and the speaker ascertains "are you with me so far?". Gussenhoven (2004, 297) mentions that this kind of meaning appears to be less intuitive with questions but may occur. We suggest, that a yes-no-question conveying this kind of meaning simply combines the speaker's ascertainment, "are you following me", with the content of the question. What additionally coincides with this kind of contour is an expectation of the answer by the speaker. This is because the speaker knows what he is talking about since he has selected the kind of information from the background (cf. Kiigler, 2003a, 2004a). For the simple rise (L*H H%), we propose that it conveys a meaning of RELEVANCE T E S T I N G . TO illustrate this, consider the passage (7). The two speakers have recently discovered that some of the symbols of their maps differ, in particular that the symbols lamb and caravan are at different places on the maps (cf. Appendix A). In the passage in (7), the instruction giver, speaker A, is about to explain the part of the route that is between the symbol that appears to be exchanged and the next symbol blaue / neue Arena 'blue / new arena'.

162

(7)

1

A:

jetz gange m a now we're going

ähm

2

uhm 3

bis zur mitte (.) until the mid (.) of d v o n dem

4

von

d of the 5

bei dlr=isch=s ä LAMM for you it's a lamb bei m i r η W0HNw[age L*H Η'/.

6

I have a caravan 7

B:

[m=hm •uh=m

The target sentence in line 6 is a conclusion of the mismatch of symbols that has been discovered before. However, it conveys a meaning that might be paraphrased as "am I right that your lamb is my caravan?". In this sense, the speaker does not commit himself as to wether the fact is true and already established in their mutually believed background. Thus, the speaker tests the relevance of the proposition, a meaning which Gussenhoven (1984) has analysed as RELEVANCE TESTING in British English. The analysis of a testing function of the sentence in line 6 is supported by the instruction receivers' confirmation in line 7 "m=hm", which is an expression of "yes". Given that the sentence in line 5 is realised with a simple rise, we conclude that the simple rise in SwabG conveys a meaning of RELEVANCE TESTING. If the sentence in lines 5 and 6 were to conclusively add the information to the conversational background of the speaker and hearer that the lamb is exchanged with the caravan, the speaker would have chosen the rise-fall (cf. above). Instead, the choice of the simple rise in this situation signals that the speaker does not commit himself as to whether or not the fact is true and already established in their mutually believed background. Thus, the speaker tests the relevance of the proposition. The meaning of RELEVANCE TESTING might be more intuitive in the case of a question (our analysis of (7) shows, it applies to declaratives). Carrying over this analysis to a question, the testing function becomes obvious, cf. (8). (8)

hOsch du da so ä aREna L*H L*H H% 'do you have a kind of arena t h e r e ? '

The instruction giver tests the relevance of the symbol Arena 'arena' in order to check whether both of them agree on the existence and position of the symbol. If they do, the instruction giver may successfully fulfill the task of further describing the route on the map. Thus, the meaning of the question in (8) may be paraphrased as "what about the symbol arena, do we agree on its existence and position or not". The instruction giver needs to update the conversational background with regards to that symbol.

163 For the modification, i.e. L-suffixation (+L), we propose a meaning of SIGNIFICANT. We have argued in chapter 4, that this modification may only apply to the rise-fall. We illustrate an example of the modified rise-fall in the passage (9). The instruction giver, speaker A, is describing the route around the symbol 'angel' and further to the next symbol 'humming dragon fly'. In line 10, the instruction receiver, speaker Β, is interrupting the route description and is elucidating the momentary position of her route drawing in order to explicitly asking for details of the route in line 11 and 12. A:

dann kommt jetzt da der ENgel then now comes the angel da simma jetz unterhalb davo there we are below of it

3

B:

4

A:

ja yes von dem ENgel of the angel dann laufe m a jetz oimal im=ä then now we are moving in=uh im=ä großzügige böge um den engel rum in a large-scale curve around the angel nach Obe upward richtung summende liBELle direction humming dragon fly ich bin jetz zwischen den FLÜgeln vom engel I am between the wings of the angel geh ich dann jetz scho zur summenden libelle rüber am I already going over to the humming dragon fly oder no um de engel drum[rum or still around the angel

1 2

5 6 7 8 9 10 11

B:

Given the description of the route around the angel, the instruction receiver wants the instruction giver to recognize her position exactly. The instruction giver shall significantly add the position on the map to the background. Based on this 'background update' which refers to a particular element of the symbol 'angel' which already is background, the instruction receiver may ask her further questions about the route. The meaning that the sentence in line 9 conveys may be paraphrased as "note that I am at the wings of the angel and add this position to our background". This significant notation is what Gussenhoven (1984, 218) has analysed as "very significant, non-routine", to which he is referring to the modification DELAY, or L-prefixation (Gussenhoven, 2004). Given that the sentence in line 9 is realised with a modified rise-fall, we conclude that the modified rise-fall in SwabG conveys a meaning of SIGNIFICANT ADDITION.

In (9), the narrow focus of the item FLUgeln 'wings' correlates with the intonational meaning of SIGNIFICANT ADDITION, expressing "It is this particular detail that I want you to significantly add to the background". In general, we assume that the modification adds the nuance of 'significance' to the base meaning, thus the modification being a morpheme, that however depends on the base morpheme.

164

Summary We propose that each nuclear contour that we have analysed in SwabG (cf. chapter 4) conveys its distinct meaning. We claim that the rise-fall conveys a meaning of A D D I T I O N , the rise-fall-rise SELECTION, and the simple rise RELEVANCE TESTING. Further, we claim that the modification L-suffixation adds an expression of 'significance' to the meaning of a base contour. Table 7.3 summarizes the SwabG contours and their respective meaning. Table 7.3: Summary of the meaning of nuclear contours in SwabG.

Meaning ADDITION SELECTION RELEVANCE TESTING SIGNIFICANT

Contour L*H L% L*HL H% L*H H% +L

In the analysis of SwabG intonation, we have noticed that nuclear contours generally start with an accentual L* tone followed by a Η trailing tone, i.e. a rise. The phonological differentiation between contours arises in the postnuclear part. With respect to meaning, we assume that the whole contour attributes to the interpretation of an intonation phrase. The crucial part that contributes to the meaning, however, appears to be the postnuclear part. If we consider the SwabG contours and the meanings they convey (cf. Table 7.3), we observe significant differences to the analysis of British English (Gussenhoven, 1984), where a simple fall, a fall-rise and a simple rise express meanings of A D D I T I O N , SELECTION and RELEVANCE TESTING, respectively. We will discuss this below comparing the SwabG analysis with that of SaxG.

7.3.2

The meaning of Upper Saxon German nuclear contours

In this section, we will analyse the meaning of SaxG nuclear contours according to the contour based approach (Gussenhoven, 1984). According to our analysis in chapter 5, in SaxG, we distinguish between three base contours and the modification L-prefixation that affects two of the base contours in our speech data. For each separate contour, we will analyse the meaning that the contour conveys. The analysis provided here attempts to illustrate our proposal how the different meanings might be correlated with which contours. We propose that the simple fall conveys a meaning of A D D I T I O N , the fall-rise SELECTION and the simple rise RELEVANCE TESTING. The modification L-prefixation adds a nuance of 'significance' to the base contours, which is equivalent to our analysis of SwabG and to British English (Gussenhoven, 1984). For the simple fall (H*L), we propose that it conveys a meaning of A D D I T I O N . TO illustrate this, consider the passage (10). The two interlocutors are discussing the symbol 'caravan' on the map, and the instruction giver, speaker A, has previously mentioned to draw the route to the caravan (line 1). The instruction receiver agrees with the instruction, and confirms the instruction in lines and 7.

165 (10)

1

Α:

und dann geht=s wieder grade nach obe[n and then it's continuing

straight up

((...)) 2

Α:

3

B:

und dOrt steht jetz η WOHNwagen and there is now a caravan aha ok

4

A:

der sieht aus wie ener aus holland that one looks like one from

5

B:

Holland

ja der steht bei mir ganz woandersch yes that's at a completely different

6

aber isch lof (.)

location

mach e mal den weg jetz hier rum

but I'll go (.) make now the route around —> 7

A:

here

bis zum WOHNwagen H*

H*L

until the caravan

From the context, it is obvious that the symbol 'caravan' and the instruction 'drawing the route to the caravan' are given, yet the information that the instruction giver communicates is that she agrees with the instruction. The sentence in line 7 thus conveys a meaning that we could paraphrase as "please add to the background that I agree with your instruction". This meaning is what Gussenhoven (1984) has analysed as ADDITION in British English. Given that the sentence in (10), line 7, is realised with the simple fall contour, we conclude that the simple fall in SaxG conveys the meaning of A D D I T I O N . 1 For the fall-rise (H*L H%), we propose that it conveys a meaning of S E L E C T I O N . To illustrate this, consider the passage (11). The speakers are discussing the symbol LAMM 'lamb', which only appears on the instruction receivers' map (cf. appendix A). The instruction receiver refers to the lamb in line 1, where the instruction giver (speaker A) has a caravan. Since the symbols differ here, the instruction giver wants to verify which symbol they are talking about or whether the instruction receiver might have taken a different route. In line 3, the instruction giver asks for that symbol, and the instruction receiver repeats that information in line 4. (11)

I B :

gut:

da hab=ich η LAM [Μ

well there I have a lamb 2

A:

[(unverständlich) (unintelligible)

3

WAS hast du da what do you have there

—• 4

B:

en LAMM H*L H'/. a lamb

1

Kohler (1991b,a) draws a distinction between an early and a medial peak in falling accents in Standard German: He claims that the early peak correlates with a meaning of 'established' or 'given'. Thus, we observe a structural and functional similarity between the SaxG simple fall in (10) and the one in Standard German. However, the structurally identical simple fall in SaxG also occurs in sentences that provide new information. Thus, the similarity between SaxG and Standard German appears not to be straightforward.

166 From the context, it is obvious that the symbol is given. In particular it has just been introduced to the conversation. The repetition in line 4 thus conveys a meaning, which we could paraphrase as "I remind you that I just introduced the symbol 'lamb' to our background". This meaning is what Gussenhoven (1984) has analysed as SELECTION in British English. Given that the answer in line 4 is realised with the fall-rise, we conclude that the fall-rise in SaxG conveys the meaning of SELECTION.

For the simple rise ( L * H H % ) , we propose that it conveys a meaning of RELEVANCE The two interlocutors are discussing several symbols of the map task, and in line 9, the instruction receiver, speaker A, asks about the symbol 'fisherman', whether the instruction giver also has this symbol since he has yet not referred to it. TESTING. TO illustrate this, consider the passage (12).

(12)

1

B:

2 3 4 5

A:

6 7

B:

8 —> 9

A:

10 Β:

un=daNEben and next to it IST bei mir I have m uh lahmender MAULesel laming mule ja yes den hab isch AUCH d[a that one I also have there [jut good das [(unverständlich) that [(unintelligible) [hast du noch en ANgler dabei L*H H'/. do you have a fisherman there ne isch hab keen ANGler dabei nix no I don't have a fisherman no

With the yes-no-question in line 9, the instruction receiver intends to ask whether the symbol 'fisherman' is relevant for the task. The speaker does not commit himself to the relevance of the symbol, i.e. whether the symbol 'fisherman' should be put into the background. So the paraphrase of line 9 would be "I ask you if fisherman is background". This meaning is what Gussenhoven (1984) has analysed as RELEVANCE TESTING. Given that the question in line 9 is realised with the simple rise, we conclude that the simple rise conveys the meaning of RELEVANCE TESTING. For questions, this meaning is apparently very intuitive. The rise-fall ( L + H * L ) and the rise-fall-rise ( L + H * L H%) are analysed as modified contours in SaxG (cf. chapter 5). For the modification, we propose that a meaning of "significance" is added to the meaning of the respective base contour. Consider the passage in (13), where we first analyse an example of a rise-fall. The instruction giver, speaker A, has noticed that the maps differ again at the position where the speakers

167

have just arrived. He states in line 1 his assumption that there must be a symbol at the instruction receiver's map. (13)

1

A:

—> 2

B:

du hast doch was genau auf der andern sei[te you must have something exactly on the other side [n LAMM hab=ich da oben links in der ecke L+H* L I have a lamb up there left in the corner

The instruction giver is requesting Β to inform him which symbol he has on his map. The request focuses on a particular symbol. The instruction receiver answers the request in line 2, with a narrow focus on the requested constituent LAMM 'lamb'. The meaning that is conveyed may be paraphrased as "I tell you that this particular symbol lamb and no other symbol in this particular position now is background". In general, we assume that the modification adds the nuance of 'significance' to the base meaning, thus the modification is a morpheme (cf. Gussenhoven, 1984, for a similar meaning in British English); in the case of the rise-fall, the meaning of 'significant' supplements the base meaning ADDITION. The meaning of 'significant' correlates well with the function of narrow focus. If we consider the rise-fall-rise, we also observe the supplementary meaning of 'significant', which is combined with the meaning of SELECTION. In ( 1 4 ) , we present an example in which the speakers are discussing the symbol 'windmill', which does not exist on the instruction receiver's map (speaker B). When the instruction giver, speaker A, notices that Β does not find a windmill on her map, he characterizes the windmill more exactly in line 4 being a rotating windmill. Prom the context, we observe that line 4 contains a narrow focus on DREhende 'rotating'. The symbol 'windmill' has been introduced to the conversation before (line 1), and in line 4, it is selected from the background, but with a particular focus on its property 'rotating'. The sentence in line 4 could be paraphrased as "I remind you that it is a particular windmill we are talking about". Again, the narrow focus correlates well with the modification. The whole meaning that the sentence in line 4 conveys is 'significant S E L E C T I O N ' . (14)

1

A:

bis du zu der WINDmühle kommst

2

B:

until you reach the m: uh

3 —> 4

5

A:

windmill

die gibt=s keine WINDmühle that does not is no windmill eine DREhende Windmühle L+HL H'/. a rotating luindmill gibt=s NICH oder wie doesn't exist or what

Given that the modification is tonally identical in (13) and (14), we conclude that it is a matter of the same modification which we analyse as L-prefixation following Gussenhoven (1984, 2004). The modification affects the whole nuclear contour in that

168

it is shifted rightwards and a leading low tonal target is inserted resulting in a salient rise to the accentual high tone. Summary To sum up, we propose that each basic nuclear contour that we have analysed in SaxG (cf. chapter 5) conveys its distinct meaning. We claim that the simple fall conveys a meaning of A D D I T I O N , the fall-rise SELECTION, and the simple rise RELEVANCE TESTING. Further, we claim that the modification L-prefixation adds an expression of 'significance' to the meaning of a base contour. Table 7.4 summarizes the SaxG contours and their respective meaning.

Table 7.4: Summary of the meaning of nuclear contours in SaxG.

Meaning ADDITION SELECTION RELEVANCE TESTING SIGNIFICANT

Contour H*L H*L H% L*H H% L+

If we compare the SaxG analysis (cf. Table 7.4) with the analysis of British English by Gussenhoven (1984), we observe striking similarities, i.e. the three basic meanings are expressed by exactly the same contours (though the phonetic realisation of these contours differ, cf. the results of Grabe, 1998a, for English as compared to our analysis of SaxG in chapters 5 and 6). As we have mentioned for the analysis of SwabG intonational meaning, the claims we make here are far from being exhaustive or comprehensive. The analysis serves as a starting point for future research.

7.4

Discussion - Two arguments in favour of a contour approach

Our analysis of intonational meaning in SwabG and SaxG has shown that we can apply the contour based model of intonational meaning (Gussenhoven, 1984). For each dialect, we have analysed the meaning of the respective base contours. We propose that each base contour conveys a distinct meaning. Following Gussenhoven, we have further argued that in both dialects, a modification of the base contours adds a distinct meaning to the meaning of the base contours. Table 7.5 summarizes our analysis and compares which meanings are expressed by means of which nuclear contour. As can be seen, SwabG and SaxG differ in the expression of the different meanings. A D D I T I O N is expressed by means of a rise-fall in SwabG and by means of a simple fall in SaxG. S E L E C T I O N is expressed by means of a rise-fall-rise in SwabG and by means of a rise-fall in SaxG. Yet, RELEVANCE TESTING is expressed by means of a simple rise in both dialects. With respect to the modification we observe small but significant differences. In SwabG, a modification results in a tonal suffixation (+L),

169

Table 7.5: Comparison of the meaning of nuclear contours between SwabG, SaxG and British English.

Meaning ADDITION SELECTION RELEVANCE TESTING SIGNIFICANT α

SwabG L*H L% L*HL H% L*H H% +L

SaxG H*L H*L H% L*H H% L+

Br. English α H*L L% H*L H% H* H% / L*H H% L*

British English according to Gussenhoven (2004, 296ff)

whereas in SaxG, it is a matter of a tonal prefixation (L+); yet in both dialects, it is a low tone that is the tonal expression of this modification. We assume that it is the same grammatical mechanism underlyingly, i.e. tonal affixation, which results in prefixation in one dialect and suffixation in the other. If we compare the dialects with British English, we observe that SaxG is by and large identical to British English whereas SwabG differs. 2 A minimal difference between SaxG and British English is a notational one; in case of A D D I T I O N , Gussenhoven represents a simple fall with a low boundary tone instead of analysing it with an unspecified boundary (cf. the proposal for English put forward by Grabe, 1998a). The second difference concerns the expression of RELEVANCE TESTING, where Gussenhoven assumes two contours. An explanation for this might be that English intonation appears to be "fairly complex" (Gussenhoven, 2004, 313) compared to other intonation languages, thus employing more intonational options to express a certain kind of meaning. If we compare SwabG with British English, we observe the same differences as between SwabG and SaxG before, yet the comparison shows a striking similarity between these three systems. If we focus on the postaccentual tones in SwabG, i.e. ignore the accentual L* tone, SwabG intonation appears to be fairly similar to SaxG and British English; for the expression of A D D I T I O N it is a tonal sequence of HL, a fall; for the expression of SELECTION, it is a tonal sequence of HLH, a fall-rise; and for the expression of RELEVANCE TESTING, it is a tonal sequence of H H , a rise. As we have pointed out in our phonological analysis of SwabG in chapter 4, the crucial phonological distinctions arise in the postnuclear part of the contours. One could thus assume that in SwabG intonational meaning is expressed in the postnuclear part of a contour. These striking similarities between SwabG, SaxG and British English appear to favour a contour based approach to intonational meaning. We have to acknowledge that yet another model of intonational meaning, the compositional model, has been proposed. We will now briefly provide two arguments why we prefer a contour based approach to intonational meaning, and thus to the phonological analysis in general. In a compositional analysis of a SaxG and a SwabG example, we attempt to show that the compositional framework will fail to analyse the meaning of intonation in the two dialects appropriately. 2

To our knowledge, there is no comprehensive analysis of intonational meaning for Standard German which we could relate our analysis to.

170 Recall the SaxG example in (14) that we reproduce here with its nuclear contour in (15). If we analyse the contour in (15) according to the compositional model (Pierrehumbert and Hirschberg, 1990), we would first have to re-analyse the tonal make-up of the contour since the Pierrehumbert framework is based on the assumption of phrase accents (cf. chapter 1, section 1.3). For the purposes of illustration, we will ignore our objections against such an analysis (cf. chapter 5). The original contour of (14) (L+H*L H%) is thus re-analysed as L+H* L - H% in (15).

A:

eine DREHende Windmühle L+H* LH% 'a rotating windmill'

As far as the pitch accent (L+H*) is concerned, the analysis of Pierrehumbert and Hirschberg (1990, 296) would apply here as well, since the L+H* accent, in Pierrehumbert and Hirschberg's terminology, 'opens a scale' expressing a contrast or correction (cf. Table 7.1 above). This is in line with our analysis of narrow focus in chapter 5 and the meaning of L-prefixation (cf. above). The problem is how to approach a meaningful analysis of the phrase accent. According to Pierrehumbert and Hirschberg, the phrase accent takes scope over an intermediate phrase, and the low phrase accent expresses a meaning of "separation of the current phrase from a subsequent phrase" (Pierrehumbert and Hirschberg, 1990, 302). In (15), there is only one phrase to take scope over. In case of a narrow focus, however, one would think of a tone taking only scope over the narrow focused constituent. The interpretation of the phrase tone as taking scope over the whole phrase appears meaningless. As a consequence of the assumption of a phrase accent, a further complication with the compositional model arises. Given that our tonal analysis of broad and narrow focus in chapter 5 holds, then broad focus is signaled by a simple falling accent (H*L) and narrow focus by a rising-falling accent (L+H*L). From a phonological point of view, one would wish to have a structural similarity between the two expressions for focus. If we consider the phonological analysis that the contour based and the compositional approach would predict, we observe that within the compositional approach this structural similarity is not present (cf. the two tonal options in (16)). (16)

Pierrehumbert and Hirschberg (1990) narrow focus: L+H* L - H% broad focus: H*L L - H%

Gussenhoven (1984) L+H*L H% H*L H%

If comparing the tonal labelling for broad and narrow focus, the Gussenhoven style shows a structural similarity between the two where narrow focus is tonally marked by an additional low tonal prefix, other tonal labels being the same. The meaning of the unmodified contour (SELECTION), i.e. a broad focus contour, still is valid for the modified contour, i.e. the narrow focus contour, with the addition of 'significance' (cf. above). The three-tonal pitch accent appears to be marked, however. The markedness of the pitch accent correlates well with the functional markedness, assuming that narrow focus is the marked case.

171

There is no structural similarity in the Pierrehumbert and Hirschberg (1990) way of labelling. As our phonetic analysis of the simple falling accent shows (cf. chapter 6), the low tone cannot be analysed as a phrase tone that aligns on postnuclear prominent syllables (in the sense of Grice et al., 2000), but instead belongs to the pitch accent. Thus the contour cannot be analysed as H* L— H%. Assuming a low phrase tone, this tone must be labelled even in cases where there is no change in pitch direction so that this tone becomes real on the surface (cf. Grice, 1995, for a detailed stepby-step analogy argument based on relevant structural turning points in pitch for a rise-fall-rise as compared to a rise-fall contour). In case of broad focus, two low tones appear on the surface, the first belonging to the simple falling pitch accent and the second being the phrase tone. However, there does not appear to be a docking site for the low phrase tone. In case of narrow focus, since the pitch contour exhibits a rise-fall-rise there is no extra need for a falling pitch accent. It is the pitch accent that differs between the two closely related contours. To close this with Pierrehumbert and Hirschberg (1990, 285), "Tunes that share certain tonal features seem intuitively to share some aspects of meaning", we fully agree with them. Although Pierrehumbert and Hirschberg intend to refer to tunes that share a low phrase and high boundary tone for instance, the authors's argument holds for our (and originally for Gussenhoven's) analysis, where a modified contour maintains the tonal structure of the nuclear base contour. From our analysis of SwabG intonation, we gain a second objection to the compositional model. The compositional model is based on the assumption that each tone of an intonational tune contributes to the meaning that is conveyed by the tune. Given our analysis of SwabG nuclear contours in chapter 4, the claim that each tone contributes to the meaning of a nuclear contour would make the wrong predictions, since every contour in SwabG starts with an identical tonal sequence that would mask any individual tonal contribution. The crucial distinctions between the contours in SwabG arise first after the accentual low tone, yet the compositional approach would claim that the accentual low tone contributes to the meaning of the nuclear tune. (If the accentual low tone were to contribute to the tune meaning, one would certainly have to re-analyse the system of meanings that individual tones contribute to the tune meaning, e.g. Table 7.1 above, since the meaning of L* in that system refers to 'salient without prediction', which is not true for SwabG; rather, a meaning that H* contributes might be applicable for the SwabG L*.) The contour based model does however appear to capture the meanings conveyed by SwabG intonation patterns. As we have argued above, the contour based model captures the SwabG meanings that nuclear contours express. This model may abstract away from redundant information such as the accentual low tone in all nuclear contours. And as we have indicated above, if abstracting away from the individual tones of a nuclear contour, striking similarities between the general shape of contours between SwabG, SaxG and British English arise. Future research, though, would have to prove our tentative claims that we have presented here. The analysis and discussion we have provided here does however reveal a fairly thorough picture of the intonational meaning and its distinct tonal expression in SwabG and SaxG, which might contribute to further debates about the meaning of intonation.

8

Summary and Conclusion

8.1

Introduction

This study is concerned with dialectal variation in intonation. In the introduction (chapter 1), we have outlined that research in intonational variation contributes to advance the theory of intonational phonology (cf. among others Grabe, 2002; Gilles and Peters, 2004a; Gussenhoven, 2004). Early dialectological work in German has always emphasized that the speech melody is recognized as a particular dialect property (e.g. Bremer, 1893), yet these hints have long been impressionistic in nature mainly due to the lack of adequate theoretical access for the study of intonation (e.g. Heike, 1983; Auer et al., 2000; Wiesinger, 1994, 19). However, the framework of intonational phonology briefly introduced in chapter 1, on which the present study is based, is a powerful theory to capture the phonology and phonetics of intonation in a language. It is a widely acknowledged system that is flexible enough to distinguish between different levels of intonational description. This is crucial for comparative (Grabe, 1998a) as well as typological (e.g. Jun, 2005) purposes. This study pursues three aims. First, we intend to provide a detailed and comprehensive analysis of intonation patterns of Swabian German (SwabG) and Upper Saxon German (SaxG). Our analysis is based on two assumptions: (i) we assume that the two dialects exhibit a specific distinct intonation; (ii) we assume that the speakers we have recorded represent the particular dialect variety. Our analysis appears to prove these assumptions and we conclude that SwabG and SaxG exhibit a distinct tonal grammar each (for a summary see section 8.2 below). The second aim is to perform a comparative analysis of the intonation of the two dialects in terms of phonetic realisational aspects. In the phonetic analysis, we intend to show that measurements of tonal alignment support our phonological analysis. Given the invariance hypothesis of tonal alignment (cf. chapter 6 and, among others, Ladd et al., 1999, 2000), we show that the individual tonal categories differ significantly from one another. On the basis of the alignment measurements in combination with measurements of pitch excursion we propose a two-dimensional model that captures the variation in accent realisation across language varieties quantitatively. The third aim is to touch upon aspects of intonational meaning. Based on the assumption that nuclear contours convey intonational meaning, we attempt to provide a provisional analysis of SwabG and SaxG nuclear contours applying the framework of Gussenhoven (1984). Our results are provisional since we do not pretend an exhaustive and comprehensive analysis but we only present examples to illustrate the general line of our proposal. An intensive analysis of intonational meaning would be beyond the scope of the present study, yet, our results appear to provide a challenging starting point for future research in intonational meaning. In chapter 2, we presented the empirical basis of the present study. We briefly provided the reasons for the analysis of a South-Western High German dialect (SwabG) and an East Middle German one (SaxG). A short characteristic of geographical and

173 salient segmental dialect properties was given. Moreover, we illustrated the design of the speech corpus, speech elicitation, background information of the informants and information about the speech material used for analysis. In chapter 3 we discussed previous work of German intonation. We started the discussion with the pioneers of intonation research (in particular Klinghardt, 1925) and continued with studies carried out in the framework of the theory of intonational phonology. Within this framework, we identified two major schools of intonation research that we colloquially referred to as the Pierrehumbert and the Gussenhoven style; note that Gussenhoven's theory of intonation research is a development from the originally proposed autosegmental-metrical theory of intonation (Pierrehumbert, 1980) still drawing on core assumptions of the original proposal. German intonation has been modeled according to these two styles, thus contradictory views on German intonation exist (e.g. Fery, 1993; Grabe, 1998b; Grice and Baumann, 2002). Prom our survey of the two distinct approaches to German we concluded to follow the Gussenhoven style with additions for German intonation based on Fery (1993) and Grabe (1998a). In particular, we assume a system of intonation research that (i) (ii) (iii) (iv) (v) (vi)

acknowledges two levels of tonal representation; assumes only complex pitch accents underlyingly; allows tonally unspecified intonation phrase boundaries; considers nuclear contours as morphemes; rejects the phrase accent as a tonal category for analysis; follows the taxonomy of cross-linguistic differences in intonation proposed by Ladd (1996, 119ff);

Chapters 4 and 5, respectively, presented a detailed analysis of SwabG and SaxG intonation patterns that we will compare and summarize in section 8.2. Chapter 6 provided a comparison of realisational aspects of SwabG and SaxG intonation that we will summarize in section 8.3. And finally, chapter 7 provided an analysis of the meaning of the nuclear contours we proposed for SwabG and SaxG respectively, which we will summarize in section 8.4.

8.2

Intonational phonology of Swabian and Upper Saxon German

Surveying previous research on Southern German intonation revealed a striking consensus about the fact that neutral accentuation in Southern German intonation is expressed by means of rising-falling intonation (cf. Table 4.1 in chapter 4). Our analysis supports this view. Further, we observed that SwabG intonation is characterized by only employing nuclear rising accents; we claimed that the tonal grammar of SwabG does not contain nuclear falling accents (cf. (1) below). The tonal grammar of SwabG is considerably distinct from Standard German, yet it also shows obvious similarities to other Southern German varieties, in particular if compared to the closely related variety of Freiburg Alemannic German (Gilles, 2005; Peters, 2006).

174 For SaxG intonation, contradictory claims have been made as regarding accentuation. Some researchers claimed that SaxG intonation is characterized by low accentuation (e.g. von Essen, 1940), whereas others claimed the opposite, i.e. high accentuation (Gericke, 1963). According to our analysis of Leipzig SaxG, we concluded that both views can be correct; the crucial difference between the contradictory claims may be the result of an examination of different accents or, perhaps, that the same accent (presumably the modified simple fall, cf. below) has been interpreted differently. No matter which explanation may hold, our comprehensive analysis made a distinction between two closely related accents. The assumption of two different nuclear falling contours, i.e. a simple fall (H*L) and a modified simple fall ( L + H * L ) , appears to be dialect-specific, at least compared to Standard German. Comparing the tonal inventories of SwabG and SaxG with each other we observed obvious differences (cf. (1) and (2) below). However, our analysis also showed clear similarities between the two dialects. In SwabG and SaxG, we distinguished between three base contours that can be affected by a tonal modification. The difference between the dialects arises in the tonal composition of these contours. We claimed that the base contours in SwabG intonation are the rise-fall (L*H L%), the simple rise (L*H H%), and the rise-fall-rise (L*HL H%), respectively (cf. (1)). Additionally, we proposed a modified contour to exist, which we colloquially referred to as the rise-fall-low-plateau ( L * H + L ) . The modification results in a structurally similar contour if compared to the base contour, in this case the rise-fall (L*H L%). Based on Gussenhoven (2004) we referred to the modification as to L-suffixation. Moreover, due to structural reasons, we argued that the modification must not affect the simple rise and the rise-fall-rise since contours would emerge on the surface that would be structurally ambiguous with the base contours. We argued that the Obligatory Contour Principle (OCP) applies in SwabG which accounts for this constraint on the modification. The tonal grammar of Swabian German is summarized in (1) (1)

A tonal grammar of Swabian German.

OCP (2)

A tonal grammar of Upper Saxon German.

For SaxG, we claimed that the base contours are the simple fall (H*L), the simple rise (L*H H%), and the fall-rise (H*L H%), respectively (cf. (2)). Additionally, we proposed a tonal modification of L-prefixation (cf. Gussenhoven, 2004) that results in at least two further contours, the rise-fall ( L + H * L ) and the rise-fall-rise ( L + H * L H%). The tonal grammar of Upper Saxon German summarized in (2) predicts two further nuclear contours, a rise without a high boundary tone, and modified rising

175

contour. We observed the former contour (L*H) in our data and argued that it is a phonetic variant of the simple rise. The latter contour (L+L*H H%) did not occur in our speech data, yet we expect this pattern to exist in SaxG. Considering the tonal modification that applies in SwabG and SaxG, we observed first that it is a matter of the same functional process, i.e. according to our analysis of intonational meaning in chapter 7 the modification adds a nuance of 'significance' to the meaning of the base contour. With respect to tonal implementation of the modification, we observed that the two dialects differ. In SwabG, a low tone is added after the nuclear accent, whereas in SaxG it is before. Since this modification is similar to one of the modifications proposed for British English (Gussenhoven, 1984, 2004), we borrowed the term L-prefixation from Gussenhoven to refer to the SaxG case. For SwabG, however, we referred to the modification as to L-suffixation. The similarity of the modification in form and function between the dialects led us to assume that this modification should be referred to as L-affixation, which causes an L prefix in one dialect, and an L suffix in the other. Finally, considering the proposed boundary specification for SwabG (1) and SaxG (2), we observed that SwabG exhibits a two-way distinction (H% vs. L%), whereas SaxG only has one, a high boundary tone (H%). A tonally unspecified intonation phrase boundary was accounted for assuming that the trailing tone of the nuclear accent spreads until the phrase boundary (Fery, 1993; Grabe, 1998a). The need for a low boundary tone in SwabG arose from the base contour rise-fall (L*H L%); in the rise-fall, the falling part is unambiguously aligned with the intonation phrase boundary. Thus, our data appear to support an assumption put forward by Grabe (1998a) that the tonal specification of boundary tone is dialect- or language-specific. Having established that the two dialects differ in their tonal inventories, the question arises, what consequences do these differences have with respect to form (phonetic implementation of tones) and function (intonational meaning). The next section (8.3) summarizes our findings regarding differences in intonational form, and section 8.4 our analysis of intonational function.

8.3

Comparative intonational phonology: Phonetic realisation

In chapter 6, we were concerned with realisational differences of intonation, i.e. a different phonetic implementation of an identical phonological category. In particular, two aspects of realisational differences, tonal alignment and pitch excursion were discussed. With regard to tonal alignment, we assumed the invariance hypothesis to hold, which predicts that the specific alignment of tones determines which tonal category is involved. Our analysis presented phonetic evidence that supported our previous phonological analysis of SwabG and SaxG intonational patterns. In particular, for SwabG, we found that the accentual low tone does not differ across contours. In contrast, for the alignment of the trailing high tone, a slightly varying pattern was observed. However, independent samples t-test comparisons confirmed the phonological analysis that assumed a structural similarity between the base contour rise-fall

176 (L*H L%) and its modified version (L*H+L). Further, t-test comparisons between the rise-fall-rise (L*HL) and the two versions of the rise-fall (L*H L% and L*H-|-L) revealed significant differences, which supported the phonological distinction between the two nuclear accents L*H and L*HL in SwabG (cf. (1) above). Similar analyses for SaxG also supported the phonological analysis. In particular, we found significant differences in Η tone alignment between the simple fall (H*L) and the modified version ( L + H * L ) . Further, we observed a significant difference in L tone alignment between the simple rise (L*H H%) and the modified simple fall ( L + H * L ) , which supported a distinction between the status of the L tone (an accentual tone vs. a leading tone). Thus, the phonological distinction between two nuclear accents and a modification expressed by a leading low tone has been phonetically proven (cf. (2) above). In a comparison of SwabG and SaxG, we proposed a two-dimensional pitch accent model that quantifies the variation of individual tonal categories cross-dialectally. Based on our alignment measurements, we combined these with an analysis of pitch excursion sizes. We concluded that these two factors appear to play an important role in capturing intonational variation across dialects.

8.4

Comparative intonational phonology: Intonational meaning

From the phonological analysis, we obtained crucial differences in the tonal inventory between SwabG and SaxG. We observed that the nuclear base contours in SwabG appear to be more complex than the ones in SaxG, and, in addition, are composed different tonally. In chapter 7, we attempted to analyse the meanings conveyed by the individual nuclear contours in order to get an impression of which nuclear contours can be interpreted as being functionally equivalent between SwabG and SaxG. Equivalent contours were summarized in Table 7.4 on page 168 in a comparison with functionally equivalent nuclear contours of British English (Gussenhoven, 2 0 0 4 ) . The analysis of meaning was based on the contour based model proposed by Gussenhoven ( 1 9 8 4 ) . The dialects differ in the expression of the meanings A D D I T I O N and S E L E C T I O N . Further, as we have mentioned above, we assumed that the modification is tonally different, but the underlying process (L-affixation) is similar between SwabG and SaxG. Comparing the individual contours with each other, we observed striking similarities, though. Focusing on the postaccentual tones in SwabG, i.e. ignoring the accentual L* tone, SwabG intonation appeared to be fairly similar to SaxG and British English; for the expression of A D D I T I O N it is a tonal sequence of HL, a fall; for the expression of S E L E C T I O N , it is a tonal sequence of HLH, a fall-rise; and for the expression of R E L E V A N C E T E S T I N G , it is a tonal sequence of H H , a rise. As we have pointed out in our phonological analysis of SwabG, the crucial phonological distinctions arise in the postnuclear part of the contours (chapter 4). We showed that this is also true for the expression of intonational meaning.

177 Despite obvious phonological differences between the two dialects, the similarities in nuclear contours led us to assume that the contour based approach adequately accounts for intonational meaning in SwabG and SaxG. In chapter 7, we further discussed that our data could be misleadingly interpreted in a compositional account of intonational meaning (Pierrehumbert and Hirschberg, 1990). We believe that our analysis, though a provisional one, supported the contour based approach and detected drawbacks of the compositional model. Although our analysis in chapter 7 was provisional and by no means exhaustive, we believe that we provided a fairly thorough picture of the intonational meaning and its distinct tonal expression in SwabG and SaxG. We hope that our analysis contributes to continue debates about the meaning of intonation.

8.5

Conclusion

The work presented in this study contributes to a better understanding of intonational phonology. Accurate and comprehensive descriptions of dialectal variation have been called for (Gussenhoven, 2004, xviii). We believe we contribute to the growing body of thorough analyses of regional variation. We have shown that dialects differ in their phonological tonal inventory, in the phonetic realisation of tones and in the expression of intonational meaning, and we have also shown that dialects share a good amount of intonational properties, or universal principles. With respect to phonetic implementation of tones, we have proposed a two-dimensional model that captures dialectal variation in a quantitative way. Still, we are aware that our study did not aim to be a methodological one in order to systematically test such a model. Therefore, future research may consider the proposed accentimplementation model as a basis for inquiry. From the point of view of intonational meaning, we hope that our tentative analysis and provisional proposal may contribute to further debates about modeling intonational meaning. In particular, considering a broader range of data would certainly prove our claims of SwabG and SaxG intonational meaning. Considering questions, for instance, might be a challenging sort of data to account for in terms of intonational meaning. We suppose that apart from the propositional content of a question fairly subtle changes in meaning might arise if uttering one and the same question with different contours (cf. Gunlogson, 2001; Kiigler, 2003a, 2004a). Last but not least, this study has analysed dialect intonation from an acoustic point of view presenting evidence that dialects exhibit particular dialect features on the phonetic as well as on the phonological level of intonational representation. Certainly, one would like to learn more about the role of dialect intonation patterns in perception (cf. Peters et al., 2002).

A

Map Tasks

The map tasks used for this investigation are displayed on the following pages. The map with the route marked on it is displayed on the lefthand page each time; the corresponding map which lacks the route is displayed on the righthand page. In this comparison, the differences between the corresponding maps become obvious: the symbols themselves or the symbols' labels can differ, or a symbol given on one map may be missing on the other. The procedure for obtaining speech data by means of the Map Task game is presented in chapter 2, page 14. I am grateful to Kathrin Claßen at the IMS Stuttgart who provided me her map task files. In Claßen (2000), four map tasks for German are presented. Two of them, maps II and III, have been chosen for the present investigation. The maps are printed with permission by Kathrin Claßen.

180

Neue Arena Wohnwagen

Indianerinnenreservat

χ Ziel

Iranische Armee junger Juwelier

J

drehende Windmühle

summende Libelle

3

Engel

Nonnenweiher MapII_l

Figure A . l : Map task 1 (instruction giver), from Claßen (2000, map I I _ 1 ) .

junger Juwelier

Wohnwagen

summende Libelle

i

Engel

Nonnenweiher MapII_2

Figure Α.2: Map task 1 (instruction receiver), from Claßen (2000, map II

2).

182

Start Χ ι ι ι

heller Diamant

I

blühende Blumen

Mehl

Ö E

/ S

Goldene Moschee

einsame Ruine

in

Λ

Sängerinnenseminar

Ziel X \

Ulmenwälder

^

lahmender Maulesel

Figure Α.3: Map task 2 (instruction giver), from Claßen (2000, map IU

I).

183

Start Χ

A

A heller Diamant

blühende Blumen

Mehl

blühende Blumen

Baumwolle ) D

Sängerinnenseminar

Angler

Ulmenwälder

lahmender Maulesel MapIII_2

Figure Α.4: Map task 2 (instruction receiver), from Claßen (2000, map III_2).

Bibliography

Albrecht, Karl (1881): Die Leipziger Mundart. Leipzig: Arnoldische Buchhandlung. Alter, Kai (1997a): Fokusprosodie im Russischen: Phonologische und akustische Korrelate von Informationsstrukturierung. In: Uwe Junghans and Gerhild Zybatov (eds.), Formale Slavistik, Frankfurt/Main: Vervuert Verlag, pages 399-414. - (1997b): Russian Prosody: Phrasing and tonal structure. In: M. Lindseth and S. Franks (eds.), Annual Workshop on formal approaches to slavic linguistics. The Indiana meeting 1996, pages 1-16, Ann Arbor: Michigan Slavic Publications. Altmann, Hans, Anton Batliner and Wilhelm Oppenrieder (eds.) (1989): Zur Intonation von Modus und Fokus im Deutschen. Tübingen: Niemeyer. Ammon, Ulrich and Uwe Loewer (1977): Schwäbisch. Düsseldorf: Pädogigischer Verlag Schwann. Anderson, Anne H., Miles Bader, Ellen G. Bard, Elizabeth Boyle, Gwyneth Doherty, Simon Garrod, Stephen Isard, Jaqueline Kowtko, Jan McAllister, Jim Miller, Catherine Sotillo, Henry S. Thompson and Regina Weinert (1991): The HCRC Map Task Corpus. Language and Speech 34(4), 351-366. Arvaniti, Amalia, D. Robert Ladd and Ineke Mennen (1998): Stability of tonal alignment: the case of Greek prenuclear accents. Journal of Phonetics 26, 3-25. - (2000): What is a starred tone? Evidence from Greek. In: M. Broe and J. Pierrehumbert (eds.), Papers in Laboratory Phonology V, Cambridge: Cambridge Univeristy Press, pages 119-131. Atterer, Michaela and D. Robert Ladd (2004): On the phonetics and phonology of "segmental anchoring" of F0: evidence from German. Journal of Phonetics 32, 177-197. Auer, Peter (1997): Führt Dialektabbau zur Stärkung oder Schwächung der Standardvarietät? Zwei phonologische Fallstudien. In: Klaus J. Mattheier and Edgar Radtke (eds.), Standardisierung und Destandardisierung europäischer Nationalsprachen, Frankfurt: Peter Lang, pages 129-161. - (2001): 'Hoch ansetzende' Intonationskonturen in der Hamburger Regionalvarietät. In: Jürgen E. Schmidt (ed.), Neue Wege der Intonationsforschung, volume 157-158 of Special edition of Germanistische Linguistik, Hildesheim: Olms, pages 125-165. Auer, Peter, Peter Gilles, Jörg Peters and Margret Selting (2000): Intonation regionaler Varietäten des Deutschen. Vorstellung eines Forschungsprojekts. In: Dieter Stellmacher (ed.), Dialektologie zwischen Tradition und Neuansätzen. Beiträge der internationalen Dialektologentagung, Stuttgart: Steiner: Göttingen, 19.-21. Oktober 1998, pages 222-239. Bannert, Robert and Anne-Christine Bredvad-Jensen (1975): Temporal organization of Swedish tonal accents: The effect of vowel duration. Working papers 10, Phonetics Laboratory, Department of General Linguistics, Lund University, Sweden, 2-36. - (1977): Temporal organization of Swedish tonal accents: The effect of vowel duration in the Gotland dialect. Working papers 15, Phonetics Laboratory, Department of General Linguistics, Lund University, Sweden, 122-128. Barker, Geoffrey Sean (2002): Intonation Patterns in Tyrolean German: An AutosegmentalMetrical Analysis. PhD thesis, University of California, Berkeley. Barker, Marie L. (1925): A Handbook of German Intonation. Cambridge: Heffer. Bartels, Christine (1999): The intonation of English statements and questions: a compositional interpretation. New York, London: Garland. Baumann, Stefan (2006): The Intonation of Givenness. Evidence from German. Tübingen: Niemeyer.

186 Baumann, Stefan, Martine Grice and Ralf Benzmüller (2001): GToBI - a phonological system for the transcription of German intonation. In: Stanislaw Puppel and Grazyna Demenko (eds.), Prosody 2000. Speech Recognition and Synthesis, pages 21-28, Poznan: Adam Mickiewicz University, Faculty of Modern Languages and Literature. Beckman, Mary E. and Gayle Ayers-Elam (1997): Guidelines for ToBI Labelling, [http://www.ling.ohio-state.edu/research/phonetics/E_ToBI/ singer_tobi.html]. Beckman, Mary E. and Janet Pierrehumbert (1986): Intonational structure in Japanese and English. Phonology Yearbook 3, 255-309. Benzmüller, Ralf and Martine Grice (1998): The nuclear accentual fall in the intonation of Standard German. ZAS Papers in Linguistics 11, Berlin: Zentrum für Allgemeine Sprachwissenschaft, Sprachtypologie und Universalienforschung, 79-98. Bergmann, Gunter (1989): Upper Saxon. In: Charles V. J. Russ (ed.), The Dialects of Modern German. A Linguistic Survey, Stanford, CA: Stanford University Press, pages 290-312. Bergmann, Pia (2006): Regionalspezifische Intonationsverläufe im Kölnischen. Formale und funktionale Analysen steig end-fallender Konturen. PhD thesis, Universität Freiburg. Bierwisch, Manfred (1966): Regeln für die Intonation deutscher Sätze. Studia Grammatica VII, 99-201, [Untersuchungen über Akzent und Intonation im Deutschen]. Boersma, Paul and David Weenink (2005): Praat: doing phonetics by computer (Version 4.2.34) [Computer program], Retrieved from http://www.praat.org/. Bohnenberger, Karl (1928): Die Mundarten Württembergs. Eine heimatliche Sprachkunde. Stuttgart: Silberburg. Bolinger, Dwight L. (1951): Intonation: levels versus configurations. Word 7, 199-210. - (1958): A therory of pitch accent in English. Word 14(2-3), 109-149. - (1972): Accent is predictable (if you're a mind-reader). Language 48(3), 633-644. - (1978): Intonation across languages. In: J. Greenberg (ed.), Universals of Human Language, volume 2, Palo Alto: Stanford University Press, pages 471-524. - (1986): Intonation and its parts. Stanford: Stanford University Press. Braun, Angelika (1996): Zur regionalen Distribution von VOT im Deutschen. In: Angelika Braun (ed.), Untersuchungen zu Stimme und Sprache, volume 96 of ZDL-Beihefte, Stuttgart: Franz Steiner, pages 19-32. Bremer, Otto (1893): Deutsche Phonetik. Leipzig: Breitkopf L· Härtel. Brinker, Klaus (ed.) (2000): Text- und Gesprächslinguistik: ein internationales Handbuch zeitgenössischer Forschung, volume 16 of Handbücher zur Sprach- und Kommunikationswissenschaft. Berlin: de Gruyter. Bruce, Gösta (1977): Swedish Word Accents in Sentence Perspective. Travaux de L'institut de Linguistique de Lund, XII. Bruce, Gösta, Claes-Christian Eiert, Olle Engstrand, Anders Eriksson and Pär Wretling (1999): Database tools for a prosodic analysis of the Swedish dialects. In: Proceedings of FONETIK 99, Department of Linguistics, Gothenburg, Sweden, pages 49-52. Bruce, Gösta and Eva Gärding (1978): A prosodic typology for Swedish dialects. In: Eva Gärding, Gösta Bruce and Robert Bannert (eds.), Nordic Prosody, Lund: Lund University, Department of Linguistics, pages 219-228. Bruce, Gösta and Ida Thelander (2001): A pitch accent journey in southern Sweden. Working papers 49, Dept. of Linguistics, Lund University, 14-17. Büring, Daniel (1996): The 59th Bridge Accent - On the meaning of Topic and Focus. SfS-Report-05-96, Seminar für Sprachwissenschaft, Tübingen University. Chambers, Jack K. and Peter Trudgill (1998): Dialectology. Cambridge: Cambridge University Press. Chomsky, Noam and Morris Halle (1968): The sound pattern of English. New York: Harper and Row.

187 Claßen, Kathrin (2000): Map Task - Eine Version für das Deutsche. AIMS, Arbeitspapiere des Instituts für Maschinelle Sprachverarbeitung 6(2), 65-82. - (2002): Realisations of Nuclear Pitch Accents in Swabian Dialect and Parkinson's Dysarthria: a Preliminary Report. In: B. Bel and I. Marlin (eds.), Proceedings of the Speech Prosody 2002 Conference, pages 223-226, 11-13 April 2002, Aix-en-Provence: Laboratoire Parole et Langage. Cruttenden, Alan (1986): Intonation. Cambridge: Cambridge University Press. - (1995): Rises in English. In: Jack Windsor Lewis (ed.), Studies in General and English Phonetics, London: Routledge, pages 155-173. Dainora, Audra (2002): Does Intonational Meaning Come Prom Tones or Tunes? Evidence Against a Compositional Approach. In: B. Bel and I. Marlin (eds.), Proceedings of the Speech Prosody 2002 Conference, pages 235-238, 11-13 April 2002, Aix-en-Provence: Laboratoire Parole et Langage. Dalton, Martha and Ailbhe Ni Chasaide (2003): Modelling intonation in three Irish dialects. In: M. J. Sole, D. Recasens and J. Romero (eds.), Proceedings of the XVth International Congress of Phonetic Sciences, pages 1073-1076, Barcelona, Spain, 3-9 August 2003. - (2005): Tonal Alignment in Irish Dialects. Language and Speech 48, 441-464. Dilley, Laura, D. Robert Ladd and Astrid Schepman (2005): Alignment of L and Η in bitonal pitch accents: testing two hypotheses. Journal of Phonetics 33(1), 115-119. Drosdowski, Günther (ed.) (1990): DUDEN. Das Aussprachewörterbuch, volume 6. Mannheim: Dudenverlag, [3rd. edition]. Engstrand, Olle, Robert Bannert, Gösta Bruce, Claes-Christian Eiert and Anders Eriksson (1997): Phonetics and phonology of Swedish dialects around the year 2000: en forskningsplan. In: Proceedings of FONETIK 1997, PHONUM, volume 4, Department of Phonetics, Umeä University, Sweden, pages 97-100. Erikson, Y. and Margit Alstermark (1972): Fundamental frequency correlates of the grave word accent in Swedish: The effect of vowel duration. STL-QPSR 2-3, KTH, Sweden, 53-60. Essen, Otto von (1940): Melodische Bewegung der Sprechstimme in deutschen Mundarten. Forschungen und Fortschritte 16(6), 63-64. - (1964): Grundzüge der deutschen Satzintonation. Ratingen: Henn, [1st edition 1956). Fery, Caroline (1988): Rhythmische und tonale Struktur der Intonationsphrase. In: Hans Altmann (ed.), Intonationsforschungen, Tübingen: Niemeyer, pages 41-64. - (1992): Focus, Topic, and Intonation in German. Arbeitspapiere des Sonderforschungsbereichs 340, Stuttgart/Tübingen. - (1993): German Intonational Patterns. Tübingen: Niemeyer. - (2000): Phonologie des Deutschen. Eine optimalitätstheoretische Einführung. [LIP 7 and 11], Postdam: Institut für Linguistik. - (2007): The Fallacy of Invariant Phonological Correlates of Information Structural Notions. In: Interdisciplinary Studies on Information Structure (ISIS), volume 6, Department of Linguistics, Potsdam University, pages 160-181. Fery, Caroline and Laura Herbst (2004): German sentence accent revisited. In: Interdisciplinary Studies on Information Structure (ISIS), volume 1, Department of Linguistics, Potsdam University, pages 43-75. Fery, Caroline and Frank Kügler (submitted): Pitch accent scaling in German. Journal of Phonetics . Fery, Caroline and Hubert Truckenbrodt (2005): Sisterhood and tonal scaling. Studia Linguistica 59(2-3), 223-243. Fischer, Hermann (1895): Geographie der schwäbischen Mundart. Tübingen: Laupp'sche Buchhandlung.

188 Fitzpatrick, Jennifer (2000): On intonational typology. In: Peter Siemund (ed.), Methodological Issues in Language Typology. Sprachtypologie und Universalienforschung, Berlin: Akademie Verlag, pages 88-96. Fitzpatrick-Cole, Jennifer (1999): The alpine intonation of Bern Swiss German. In: John J. Ohala (ed.), Proceedings of the XIVth International Congress of Phonetic Sciences (ICPhS), pages 941-944, San Francisco. Fletcher, Janet and Jonathan Harrington (2001): High-Rising Terminals and Fall-Rise Tunes in Australian English. Phonetica 58(4), 215-229. Fox, Anthony (2000): Prosodic Features and Prosodic Structure. Oxford: Oxford University Press. Frey, Eberhardt (1975): Stuttgarter Schwäbisch. Laut- und Formenlehre eines Stuttgarter Ideolekts. Marburg: Elwert. Frota, Sonia (2000): Prosody and focus in European Portuguese. Phonological phrasing and intonation. New York: Garland. Gelbe, Theodor (1875): Die sächsische Mundart und ihr Verhältnis zur Lautverschiebung. Zweiter Jahresbericht über die Realschule zu Stollberg. Gericke, Ingeborg (1963): Die Intonation der Leipziger Umgangssprache. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikation 16(4), 337-369. Gibbon, Dafydd (1998): Intonation in German. In: Daniel Hirst and Albert DiCristo (eds.), Intonation Systems. A Survey of Twenty Languages, Cambridge: Cambridge University Press, pages 78-95. Gilles, Peter (2001a): Die Intonation final fallender Nuklei. Eine kontrastive Untersuchung zum Hamburgischen und Berlinerischen. In: Jürgen E. Schmidt (ed.), Neue Wege der Intonationsforschung, volume 157-158 of Special edition of Germanistische Linguistik, Hildesheim: Olms, pages 167-200. - (2001b): Regionale Intonation. Die Intonation der Weiterweisung im Hamburgischen und Berlinischen. Zeitschrift für Germanistische Linguistik 29, 40-69. - (2002): Untersuchungen zur regionalen Färbung der Intonation des Standarddeutschen. Diskussion eines methodischen Zugangs. In: Peter Wiesinger (ed.), Akten des X. Internationalen Germanistenkongresses, Wien 2000. Zeitwende - Die Germanistik auf dem Weg vom 20. ins 21. Jahrhundert, volume 3 of Jahrbuch für Internationale Germanistik, Frankfurt: Lang, pages 249-255. - (2005): Regionale Prosodie im Deutschen. Variabilität in der Intonation von Abschluss und Weiterweisung. Berlin/New York: de Gruyter. Gilles, Peter and Jörg Peters (2004a): Introduction: Different approaches to intonational variation. In: (Gilles and Peters, 2004b), pages 1-8. Gilles, Peter and Jörg Peters (eds.) (2004b): Regional Variation in Intonation. Tübingen: Niemeyer. Gilles, Peter, Jörg Peters, Peter Auer and Margret Selting (2001): Perzeptuelle Identifikation regional markierter Tonhöhenverläufe. Ergebnisse einer Pilotstudie zum Hamburgischen. Zeitschrift für Dialektologie und Linguistik 68(2), 155-172. Goldbeck, Thomas P. and Walter F. Sendlmeier (1988): Wechselbeziehungen zwischen Satzmodularität und Akzentuierung in satzfinaler Position bei der Realisierung von Intonationskonturen. In: Hans Altmann (ed.), Intonationsforschungen, Tübingen: Niemeyer, pages 305-321. Grabe, Esther (1998a): Comparative Intonational Phonology. English and German. PhD thesis, Nijmegen University. - (1998b): Pitch accent realization in English and German. Journal of Phonetics 26(3), 129-143.

189 - (2002): Variation adds to prosodic typology. In: B. Bel and I. Marlin (eds.), Proceedings of the Speech Prosody 2002 Conference, pages 127-132, 11-13 April 2002, Aix-en-Provence: Laboratoire Parole et Langage. - (2004): Intonational variation in urban dialects of English spoken in the British Isles. In: (Gilles and Peters, 2004b), pages 9-31. Grabe, Esther and Francis Nolan (1997-2002): Intonational Variation in the British Isles, [http://www.phon.ox.uk/-esther/ivyweb/]. Grabe, Esther and Brechtje Post (2002): Intonational Variation in English. In: B. Bel and I. Marlin (eds.), Proceedings of the Speech Prosody 2002 Conference, pages 343-346, 11-13 April 2002, Aix-en-Provence: Laboratoire Parole et Langage. Grabe, Esther, Brechtje Post, Francis Nolan and Kimberley Farrar (2000): Pitch accent realization in four varieties of Britsh English. Journal of Phonetics 28(2), 161-185. Grice, Martine (1995): The intonation of interrogation in Palermo Italian. Tübingen: Niemeyer. Grice, Martine and Stefan Baumann (2002): Deutsche Intonation und GToBI. Linguistische Berichte 191, 267-298. Grice, Martine, Stefan Baumann and Ralf Benzmüller (2005): German Intonation in Autosegmntal-Metrical Phonology. In: (.Jun, 2005), chapter 2, pages 430-458. Grice, Martine, D. Robert Ladd and Amalia Arvaniti (2000): On the place of phrase accents in intonational phonology. Phonology 17(2), 143-185. Grice, Martine, Μ. Reyelt, Ralf Benzmüller, Mayer Jörg and Anton Batliner (1996): Consistency in Transcription and Labelling of German Intonation with GToBI. In: Proceedings of the Fourth International Conference on Spoken Language Processing, volume 3, pages 1716-1719, Philadelphia, USA. Grice, Martine and Michelina Savino (1997): Can pitch accent type convey information status in yes-no-questions? In: Kai Alter, Hannes Pirker and W. Finkler (eds.), Proceedings of the ACL97 Workshop on Concept-to-Speech Generation Systems, pages 29-38, Universidad Nacional de Educatie a Distancia, Madrid, Spain. - (2003a): Map Tasks in Italian: Asking Questions about Given, Accessible, and New Information. Catalan Journal of Linguistics 2, 153 180. - (2003b): Question type and information structure in Italian. In: Proceedings of Prosodic Interface 2003, Nantes, France. Gunlogson, Christine (2001): True to Form: Rising and Falling Declaratives as Questions in English. PhD thesis, University of California, Santa Cruz. Gussenhoven, Carlos (1984): On the Grammar and Semantics of Sentence Accents. Dordrecht: For is. - (1994): Review of Caroline Fery: German Intonational Patterns. Linguistics 32(3), 555558. - (2004): The Phonology of Tone and Intonation. Cambridge: Cambridge University Press. Gussenhoven, Carlos and A. Rietveld (1992): Intonation contours, prosodic structure and preboundary lengthening. Journal of Phonetics 20, 283-303. Haag, Karl (1946): Die Grenzen des Schwäbischen in Württemberg. Stuttgart: Wissenschaftliche Verlagsgesellschaft. Haan, Judith (2001): Speaking of Questions. An Exploration of Dutch Question Intonation. PhD thesis, Netherlands Graduate School of Linguistics (LOT). Hayes, Bruce and Aditi Lahiri (1991): Bengali intonational phonology. Natural Language and Linguistic Theory 9(1), 47-96. Heike, Georg (1983): Suprasegmentale dialektspezifische Eigenschaften. Überblick und Forschungsbericht. In: Werner Besch, Ulrich Knoop, Wolfgang Putschke and Herbert Ernst Wiegand (eds.), Dialektologie. Ein Handbuch zur deutschen und allgemeinen Dialektforschung, volume 1, 2 of Handbücher zur Sprach- und Kommunikationswissenschaft, Berlin: de Gruyter, pages 1154-1169.

190 Hiller, Markus (1995): Regressive Pharyngalisierung im Stuttgarter Schwäbischen als C-VInteraktion. Linguistische Berichte 155, 33-64. Hobbs, Jerry R. (1990): The Pierrehumbert-Hirschberg Theory of Intonational Meaning Made Simple: Comments on Pierrehumbert and Hirschberg. In: Philip R. Cohen, Jerry Morgan and Martha E. Pollack (eds.), Intentions in Communication, Cambridge, MA: MIT Press, pages 313-323. House, Jill and Anne Wichman (1996): Investigating peak timing in naturally-occurring speech: from segmental constraints to discourse structure. Speech Hearing and Language: work in Progress 9, Department of Phonetics and Linguistics, University College London, [http: / / www.phon.ucl.ac.uk/home/shl9/contents.htm]. Hutterer, Claus Jürgen (1975): Die Germanischen Sprachen. Ihre Geschichte in Grundzügen. Budapest: Akademiai kiadö. IPA (1999): Handbook of the International Phonetic Association: a guide to the use of the International Phonetic Alphabet. Cambridge: University of Cambridge Press. Jackendoff, Ray S. (1972): Semantic interpretation in generative grammar. Cambridge: MIT Press. Jacobs, Joachim (1982): Neutraler und nicht-neutraler Satzakzent im Deutschen. In: Theo Vennemann (ed.), Silben, Segmente, Akzente, Tübingen: Niemeyer, pages 141-169. Jarman, Eric and Allan Cruttenden (1976): Belfast intonation and the myth of the fall. Journal of the International Phonetic Association 6(1), 4-12. Jun, Sun-Ah (2005): Prosodic Typology. In: Sun-Ah Jun (ed.), Prosodic Typology. The Phonology of Intonation and Phrasing, Oxford: Oxford University Press, pages 430-458. Kerswill, Paul (1996): Divergence and convergence of sociolinguistic structures in Norway and England. Sociolinguistica 10, 90-104. Klinghardt, Hermann (1925): Sprechmelodie und Sprechtakt. Marburg: Elwert'sche Verlagsbuchhandlung. - (1927): Übungen in deutschem Tonfall. Leipzig: Quelle und Meyer. Klinghardt, Hermann and Gertrude Klemm (1920): Übungen im englischen Tonfall. Cöthen: Schulze. Klinghardt, Hermann and P. Olbrich (1925): Französische Intonationsübungen. Leipzig: Quelle und Meyer. Kohler, Klaus J. (ed.) (1991a): Studies in German Intonation, volume 25 of Arbeitsberichte (AIPUK). Kiel: Institut für Phonetik und digitale Sprachverarbeitung. Kohler, Klaus J. (1991b): Terminal Intonation Patterns in Single-Accent Utterances of German: Phonetics, Phonology and Semantics. In: (Köhler, 1991a), pages 115-185. König, Werner (1998): dtv-Atlas Deutsche Sprache. München: dtv, [12. edition]. Kügler, Prank (2003a): Do we know the answer? Variation in yes-no-question intonation. In: Susann Fischer, Ralph Vogel and Ruben van de Vijver (eds.), Experimental studies in linguistics, volume 20 of LIP, Potsdam University, pages 9-29. - (2003b): The Influence of Vowel Quality and Syllable Structure on the Temporal Organization of Pitch Accents in Berlin German. In: Proceedings of the XVth International Congress of Phonetic Sciences, pages 1221-1224, Barcelona, Spain, 3-9 August 2003. - (2004a): Dialectal variation in question intonation in two German dialects: the case of Swabian and Upper Saxon. In: Proceedings of the second International Conference on Language Variation in Europe, pages 227-240, Uppsala University. - (2004b): The phonology and phonetics of nuclear rises in Swabian German. In: (Gilles and Peters, 2004b), pages 75-98. Kügler, Frank, Caroline Fery and Ruben van de Vijver (2003): Pitch Accent Realization in German. In: Proceedings of the XVth International Congress of Phonetic Sciences, pages 1261-1264, Barcelona, Spain, 3-9 August 2003.

191 Ladd, D. Robert (1980): The structure of intonational meaning: evidence from English. Bloomington: Indiana University Press. - (1981): On Intonational Universals. In: Terry Myers, .John Laver and John Anderson (eds.), The Cognitive Representation of Speech, [Advances in Psychology 7], Amsterdam: North Holland, pages 389-397. - (1983): Phonological features of intonational peaks. Language 59(3), 721-759. - (1993): On the theoretical status of the "the baseline" in modelling intonation. Language and Speech 36(4), 435-451. - (1994): Review of Caroline Fery, German Intonational Patterns. Linguistische Berichte 154, 489-492. - (1996): Intonational Phonology. Cambridge: CUP. - (2000): Bruce, Pierrehumbert, and the elements of intonational phonology. In: Merle Hörne (ed.), Prosody: Theory and Experiment. Studies presented to Gösta Bruce, Dordrecht: Kluwer, pages 37-50. Ladd, D. Robert, Dan Faulkner, Hanneke Faulkner and Astrid Schepman (1999): Constant segmental anchoring of F0 movements under changes in speech rate. Journal of the Acoustical Society of America 106(3), 1543-1554. Ladd, D. Robert, Ineke Mennen and Astrid Schepman (2000): Phonological conditioning of peak alignment in rising pitch accents in Dutch. Journal of the Acoustical Society of America 107(5), 2685-2696. Ladd, D. Robert and Astrid Schepman (2003): "Sagging transitions" between high pitch accents in English: experimental evidence. Journal of Phonetics 31(1), 81-112. Leben, Ronald William (1980): Suprasegmental Phonology. New York: Garland. Liberman, M. and J. Pierrehumbert (1984): Intonational invariance under changes in pitch range and length. In: Language Sound Structure: Studies in Phonology presented to Morris Halle, Cambridge: M I T Press, pages 157-233. Liberman, M. and I. Sag (1974): Prosodic form and discourse function. In: Proceedings of the Chicago Linguistics Society, volume 10, pages 416-427. Lickley, Robin, Astrid Schepman and D. Robert Ladd (2005): Alignment of "phrase accent" lows in Dutch falling-rising questions: Theoretical and methodological implications. Language and Speech 48, 157-183. Mayer, Jörg (1997): Intonation und Bedeutung. PhD thesis, Universität Stuttgart, [Arbeitspapiere des Instituts für Maschinelle Sprachverarbeitung, AIMS 3(4)]. Möhler, J. and J. Mayer (1999): A method for the analysis of prosodic registers. In: Proceedings of the 6th European Conference on Speech Communication and Technology, volume 2, pages 735-738, Budaest. Nakano, Y.I., M. Naka, Y . Horiuchi, A. Yoshino, S. Tutiya, A . Ichikawa, M. Ishizaki, M. Okada, H. Koiso and H. Suzuki (1997): Basic statistics of the Japanese Map Task Corpus (1). Spoken language understanding and discourse processing Research Notes No. SIG-SLUD-9701, 19-24, Japanese Society for Artificial Intelligence. Nolan, Francis (2003): Intonational equivalence: an experimental evaluation of pitch scales. In: Proceedings of the XVth International Congress of Phonetic Sciences, pages 771-774, Barcelona, Spain, 3-9 August 2003. Patterson, David (2000): A Linguistic Approach to Pitch Range Modelling. PhD thesis, University of Edinburgh. Peters, Jörg (1999a): The timing of nuclear high accents in German dialects. In: John J. Ohala (ed.), Proceedings of the XIVth International Congress of Phonetic Sciences (ICPhS), pages 1877-1880, San Francisco. - (1999b): Trunkierung und Kompression bei finalen Akzentsilben. Vorläufige Resultate einer Untersuchung zum Berlinerischen und Hamburgischen, MS, Potsdam University. - (2001): Steigend-fallende Konturen im Berlinerischen. Deutsche Sprache 29, 122-147.

192 - (2002): Fokus und Intonation im Hamburgischen. Linguistische Berichte 189, 27-57. - (2006): Intonation deutscher Regionalsprachen. Berlin/New York: de Gruyter. Peters, Jörg, Peter Gilles, Peter Auer and Margret Selting (2002): Identification of regional varieties by intonational cues. An experimental study on Hamburg and Berlin German. Language and Speech 45, 115-139. - (2003): Identifying regional varieties by pitch information: A comparison of two approaches. In: M. J. Solö, D. Recasens and J. Romero (eds.), Proceedings of the XVth International Congress of Phonetic Sciences, pages 1065-1068, Barcelona, Spain, 3-9 August 2003. Pierrehumbert, Janet (1980): The phonology and phonetics of English intonation. PhD thesis, MIT, Distributed by Indiana University Linguistics Club, Bloomington. Pierrehumbert, Janet and Mary Beckman (1988): Japanese Tone Structure. Cambridge, Massachusetts: MIT Press. Pierrehumbert, Janet and Julia Hirschberg (1990): The Meaning of Intonational Contours in the Interpretation of Discourse. In: Philip R. Cohen, Jerry Morgan and Martha E. Pollack (eds.), Intentions in Communication, Cambridge, MA: MIT Press, pages 271-311. Pierrehumbert, Janet and Shirley A. Steele (1989): Categories of Tonal Alignment in English. Phonetica 46, 181-196. Pike, Kenneth Lee (1945): The Intonation of American English. Ann Arbor: University of Michigan Press, [Reprinted in 1979]. Prieto, Pilar, Jan van Santen and Julia Hirschberg (1995): Tonal alignment patterns in Spanish. Journal of Phonetics 23, 429-451. Putschke, Wolfgang (1968): Ostmitteldeutsche Dialektologie. In: L. E. Schmitt (ed.), Germanische Dialektologie, Wiesbaden, pages 105-154. Reetz, Henning (1999): Artikulatorische und akustische Phonetik. Trier: Wissenschaftlicher Verlag Trier. Rietveld, Toni and Carlos Gussenhoven (1995): Aligning pitch targets in speech synthesis: effects of syllable structure. Journal of Phonetics 23, 375-385. Russ, Charles V. J. (1989): Swabian. In: Charles V. J. Russ (ed.), The Dialects of Modern German. A Linguistic Survey, Stanford, CA: Stanford University Press, pages 337-363. Schädlich, Hans-Joachim and Heinrich Eras (1970): Vergleichende Untersuchungen über die Satzintonation in deutschen Dialekten. In: Proceedings of the Sixth International Congress of Phonetic Sciences, pages 793-796. Schepman, Astrid, Robin Lickley and D. Robert Ladd (2006): Effects of vowel length and "right context" on the alignment of Dutch nuclear accents. Journal of Phonetics 34, 1-28. Schneider, Katrin and Britta Lintfert (2003): Categorical Perception of Boundary Tones in German. In: M. J. Sole, D. Recasens and J. Romero (eds.), Proceedings of the XVth International Congress of Phonetic Sciences, pages 631-634, Barcelona, Spain, 3-9 August 2003. Selkirk, E. (1984): Phonology and syntax. The relation between sound and structure. Cambridge: MIT Press. Selkirk, Elisabeth (1995): Sentence prosody: Intonation, stress, and phrasing. In: John A. Goldsmith (ed.), The Handbook of Phonological Theory, Cambridge MA/Oxford: Blackwell, pages 550-569. Selting, Margret (1993): Phonologie der Intonation. Zeitschrift für Sprachwissenschaft 11, 99-138. - (1995): Prosodie im Gespräch. Aspekte einer interaktionalen Phonologie der Konversation. Tübingen: Niemeyer. - (2002a): Dresdener Intonation: Fallbögen. InLiSt - Interaction and Linguistic Structures 29, [Universität Potsdam, http://www.uni-potsdam.de/u/inlist].

193 - (2002b): Dresdener Intonation: Treppenkonturen. InLiSt - Interaction and Linguistic Structures 28, [Universität Potsdam, http://www.uni-potsdam.de/u/inlist]. - (2003): Lists as embedded structures and the prosody of list construction as an interactional resource. InLiSt - Interaction and Linguistic Structures 35, [Universität Potsdam, http://www.uni-potsdam.de/u/inlist]. - (2004): Listen: Sequenzielle und prosodische Struktur einer kommunikativen Praktik eine Untersuchung im Rahmen der Interaktionalen Linguistik. Zeitschrift für Sprachwissenschaft 23, 1-46. Selting, Margret, Peter Auer, Birgit Barden, Jörg Bergmann, Elizabeth Couper-Kuhlen, Susanne Günthner, Christoph Meier, Uta Quasthoff, Peter Schlobinski and Susanne Uhmannm (1998): Gesprächsanalytisches Transkriptionssystem (GAT). Linguistische Berichte 173(34), 91-122. Sievers, Eduard (1912): Rhythmisch-melodische Studien. Heidelberg. Silverman, Kim, Mary E. Beckman, John Pitrelli, Mari Ostendorf, Colin Wightman, Patti Price, Janet Pierrehumbert and Julia Hirschberg (1992): ToBI: A standard for labeling English prosody. In: Proceedings of the 1992 International Conference on Spoken Language Processing, pages 867-870. Silverman, Kim E. A. and Janet Pierrehumbert (1990): The timing of prenuclear high pitch accents in English. In: John Kingston and Mary E. Beckman (eds.), Papers in Laboratory Phonology I. Between the Grammar and Physics of Speech, Cambridge: Cambridge University Press, pages 72-106. Trager, George L. and Henry Lee Smith (1951): An outline of English structure. Washington: American Council of Learned Societies, [7th reprint 1966]. Truckenbrodt, Hubert (2002): Upstep and embedded register levels. Phonology 19(1), 77120. - (2004): Final lowering in non-final position. Journal of Phonetics 32, 313-348. Truckenbrodt, Hubert and Caroline Fery (Ms): More on hierarchical organization and tonal scaling. Ms, Universities of Tübingen and Potsdam. Uhmann, Susanne (1991): Fokusphonologie: eine Analyse deutscher Intonationskonturen im Rahmen der nicht-linearen Phonologie. Tübingen: Niemeyer. Ulbrich, Christiane (2004): A comparative study of declarative intonation in Swiss and German standard varieties. In: (Gilles and Peters, 2004b), pages 99-122. van den Berg, Rob, Carlos Gussenhoven and Toni Rietveld (1992): Downstep in Dutch: implications for a model. In: Gerard J. Docherty and D. Robert Ladd (eds.), Papers in Laboratory Phonology II. Gesture, Segment, Prosody, Cambridge: Cambridge University Press, pages 335-359. Waiblinger, E. (1925): Tonfall deutscher Mundarten. Vox 10, 43-44. Warren, Paul (2005): Patterns of late rising in New Zealand English: Intonational variation or intonational change? Language Variation and Change 17, 209-230. Wichman, Anne, Jill House and Toni Rietveld (2000): Discourse constraints on F0 Peak Timing in English. In: Antonis Botinis (ed.), Intonation: Analysis, Modelling and Technology, Dordrecht: Kluwer, pages 163-184. Wiese, Richard (1996): The phonology of German. Oxford: Clarendon Press. Wiesinger, Peter (1983): Die Einteilung der deutschen Dialekte. In: Werner Besch, Ulrich Knoop, Wolfgang Putschke and Herbert Ernst Wiegand (eds.), Dialektologie. Ein Handbuch zur deutschen und allgemeinen Dialektologie, volume 1, 2 of Handbücher zur Sprachund Kommunikationswissenschaft, Berlin: de Gruyter, pages 807-900. - (1994): Zum gegenwärtigen Stand der phonetisch-phonologischen Dialektbeschreibung. In: Klaus Mattheier and Peter Wiesinger (eds.), Dialektologie des Deutschen. Forschungsstand und Entwicklungstendenzen, Tübingen: Niemeyer, pages 3-27.

194 Willis, Erik W. (2003): Prenuclear Low Tone Alignment in Dominican Spanish. In: Proceedings of the XVth International Congress of Phonetic Sciences, pages 2941-2944, Barcelona, Spain, 3-9 August 2003. Wunderlich, Dieter (1988): Der Ton macht die Melodie - Zur Phonologie der Intonation des Deutschen. In: Hans Altmann (ed.), Intonationsforschungen, Tübingen: Niemeyer, pages 1-40. Zimmermann, Gerhard (1992): Das Sächsische. Muttersprache 102, 97-113. - (1998): Die 'singende' Sprechmelodie im Deutschen. Zeitschrift für Germanistische Linguistik 26, 1-16. Zybatov, Gerhild and Grit Mehlhorn (2000): Experimental evidence for focus structure in Russian. In: Tracy H. King and Irina Sekerina (eds.), Formal approaches to slavic linguistics 8: The Philadelphia meeting 1999, Ann Arbor: Michigan Slavic Publications, pages 414-434.