Intonation and Stress Evidence from Hungarian
László Varga
Intonation and Stress
This page intentionally left blank...
41 downloads
1422 Views
1MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
Intonation and Stress Evidence from Hungarian
László Varga
Intonation and Stress
This page intentionally left blank
Intonation and Stress Evidence from Hungarian László Varga
© László Varga 2002 All rights reserved. No reproduction, copy or transmission of this publication may be made without written permission. No paragraph of this publication may be reproduced, copied or transmitted save with written permission or in accordance with the provisions of the Copyright, Designs and Patents Act 1988, or under the terms of any licence permitting limited copying issued by the Copyright Licensing Agency, 90 Tottenham Court Road, London W1T 4LP. Any person who does any unauthorised act in relation to this publication may be liable to criminal prosecution and civil claims for damages. The author has asserted his right to be identified as the author of this work in accordance with the Copyright, Designs and Patents Act 1988. First published 2002 by PALGRAVE MACMILLAN Houndmills, Basingstoke, Hampshire RG21 6XS and 175 Fifth Avenue, New York, N.Y. 10010 Companies and representatives throughout the world PALGRAVE MACMILLAN is the global academic imprint of the Palgrave Macmillan division of St. Martin’s Press, LLC and of Palgrave Macmillan Ltd. Macmillan® is a registered trademark in the United States, United Kingdom and other countries. Palgrave is a registered trademark in the European Union and other countries. ISBN 0–333–97370–4 This book is printed on paper suitable for recycling and made from fully managed and sustained forest sources. A catalogue record for this book is available from the British Library. Library of Congress Cataloging-in-Publication Data Varga, László, 1943– Intonation and stress : evidence from Hungarian / László Varga. p. cm. Includes bibliographical references and index. ISBN 0–333–97370 –4 (cloth) 1. Hungarian language—Intonation. 2. Hungarian language —Accents and accentuation. I. Title. PH2139.5 .V36 2002 494′.51115—dc21 2001058215 10 11
9 10
8 7 09 08
6 07
5 06
4 05
3 2 04 03
1 02
Printed and bound in Great Britain by Antony Rowe Ltd, Chippenham and Eastbourne
Contents ix
List of Tables
x
Acknowledgements List of Intonational Transcription Symbols
xi
Other Symbols and Abbreviations
xiii
Hungarian Letter-to-Sound Correspondences
xvi
1 Introduction 1.1 Aims and scope 1.2 Theoretical assumptions 1.2.1 Melodic prosodemes as meaningful wholes 1.2.2 Syntax–phonology mapping 1.2.3 The syntactic model chosen 1.3 Sources 1.4 Past work on Hungarian stress and intonation
1 1 4 4 5 10 12 13
PART I
INTONATION
2 Intonation, Paralanguage, Prosody 2.1 Introduction 2.2 Prosodic devices and intonation 2.2.1 Intonation as part of prosody 2.2.2 Prosodic limits 2.2.3 Paralinguistic variations 2.3 The functions of intonation 2.3.1 The types of intonational functions 2.3.2 The attitudinal functions of intonation 2.3.3 The grammatical functions of intonation 2.3.4 The primacy of melodic prosodemes 2.3.5 Other groupings of intonational functions 2.4 Summary
19 19 19 19 21 22 22 22 23 24 29 29 31
3 A Taxonomic Analysis of Hungarian Intonation 3.1 Introduction 3.2 Character contours 3.2.1 On character contours and degrees of stress
32 32 33 33
v
vi
Contents
3.2.2 3.2.3 3.2.4 3.2.5
3.3 3.4 3.5 3.6
3.7 3.8
Front-falling contours Sustained contours End-falling contours Summary of the first three groups of character contours 3.2.6 The second-type descent 3.2.7 The stylized fall 3.2.8 Minor stresses do not affect the character contours The appended contour The intonational lexicon The preparatory contour The unit of Hungarian intonation 3.6.1 The intonation phrase 3.6.2 Why the contour is not an alternative unit of intonation 3.6.3 More on the intonation phrase Downdrift and upstep Summary
34 36 38 42 42 44 47 48 50 52 54 54 56 60 61 64
4 An Autosegmental Analysis of Hungarian Intonation 4.1 Introduction 4.2 The Kálmán–Kornai analysis 4.3 The representational conventions 4.4 The basic contours 4.5 Melody formation 4.5.1 Primary derivational rules 4.5.2 Secondary derivational rules 4.6 Summary of the rules and representations 4.7 The association conventions 4.8 The rearrangement rules 4.9 Phonetic realization 4.10 Structure of the Hungarian intonation phrase 4.11 Summary
65 65 66 69 70 71 71 73 75 76 80 81 85 86
5 The Melodic Segmentation of Hungarian Utterances 5.1 Introduction 5.2 Sentence and utterance 5.3 Major tonosyntactic blocks 5.3.1 Utterances with more than one highest-ranking sentence 5.3.2 Utterances with inorganic blocks
87 87 88 90 90 93
Contents vii
5.4
5.5 5.6
PART II
Minor tonosyntactic blocks: below the highest-ranking sentence 5.4.1 Division into dominant and dependent minor blocks 5.4.2 Assignment of melodies 5.4.3 Equivalent blocks (EBs) 5.4.4 Complementary blocks (CBs) 5.4.4.1 Syntactic types of complementary blocks 5.4.4.2 Dependent contours before Type α dominant contours 5.4.4.3 Dependent contours before dominant contours other than Type α 5.4.5 Melodic blocks that can be used as both complementary and equivalent blocks 5.4.6 Summary of the melody rules in the highest-ranking sentence 5.4.7 Contour replacements 5.4.8 Pauses in the highest-ranking sentence Deviations Summary
97 97 99 100 102 102 105 111 112 115 117 121 122 124
STRESS
6 Stress in Hungarian Words, Phrases and Sentences 6.1 Introduction 6.2 The phonological stress degrees of Hungarian 6.3 Hungarian stress is not fully relational 6.4 Representation 6.5 Stress in Hungarian words 6.5.1 Stress in simple and derived lexical words 6.5.2 Stress in lexical compounds 6.5.3 Stress in function words 6.6 Stress in Hungarian phrases and sentences 6.6.1 Syntactically motivated deaccentuation 6.6.2 Other kinds of deaccentuation 6.6.3 Clash deletion: obligatory destressing 6.7 Summary
127 127 127 129 131 132 132 133 136 138 138 144 146 148
7 Rhythmical Variation in Phrasal Compounds 7.1 Introduction 7.2 The facts of rhythmical variation in Hungarian
149 149 149
viii
Contents
7.3
7.4
7.2.1 Phrasal compounds and the two kinds of rhythmical variation 7.2.2 The number of syllables between the accents 7.2.3 Parallel phenomena in English The analysis of rhythmical variation in Hungarian 7.3.1 Analyses based on different accentual strengths 7.3.1.1 The adaptation of Selkirk’s model 7.3.1.2 The adaptation of metrical adjunction theory 7.3.2 Gussenhoven’s pitch accent-based analysis 7.3.3 The split analysis Summary
149 153 157 158 158 158 166 168 170 177
8 Rhythmical Secondary Stresses 8.1 Introduction 8.2 Earlier accounts 8.3 Contour Insertion 8.3.1 The mechanism and meaning of contour insertion 8.3.2 The experiment 8.3.3 Extension 8.4 The domain of rhythmical secondary stress placement 8.5 The order of rules 8.6 Summary
178 178 179 180
9 Summary and Conclusions
196
Notes
203
References
216
Index
224
180 183 188 190 194 194
List of Tables 3.1 The system of Hungarian stress prosodemes 3.2 The intonational lexicon of Hungarian 3.3 The frequency of occurrence of meaningful contours in the corpus examined 4.1 Tones and their representations 6.1 The system of Hungarian stress prosodemes 6.2 The degree cline (illustrated from English) 7.1 The asymmetry of Trochaic Reversal and Trochaic Confirmation 8.1 Summary of the statistical analysis of mean scores
ix
33 51 52 69 129 141 152 187
Acknowledgements I would like to express my special gratitude to the Hungarian linguists László Deme, Katalin É. Kiss, Ádám Nádasdy, Mihály Péter, Péter Siptár and Tamás Szende for their comments on earlier versions of various parts of this book. My thanks are also due to participants of workshops and courses at the Eötvös Loránd University in Budapest, where parts of this book were first presented. I am also thankful to anonymous reviewers for their invaluable suggestions and comments. I would like to thank Cambridge University Press for letting me reproduce part of my article “Rhythmical Variation in Hungarian”, which was published in Phonology, 15 (1998), 227–66; and to Routledge for permission to reproduce some of the examples of my “Stylization of the Falling Tone in Hungarian Intonation”, which appeared in Jack Windsor Lewis (ed.), Studies in General and English Phonetics (1995), 278–87. I am also indebted to many scholars of the universal and comparative aspects of intonation, especially to the late Professor Dwight Bolinger, and to Professors Alan Cruttenden, Ivan Fónagy, Carlos Gussenhoven and Robert Ladd. They have all helped me to form my thoughts on the subject. At the same time, I would like to emphasize that I myself am responsible for any shortcomings that this book may have. LÁSZLÓ VARGA
Budapest
x
List of Intonational Transcription Symbols
For a detailed explanation see Chapter 3. The letter “x” in the list below represents the written form of the syllable.
(i)
Melodic and stress prosodemes
1x, [x, 7x
combined symbols for “front-falling” character contours (full fall, half fall, fall–rise) and the major-stressed syllables initiating them
2x, 4x, 5x
combined symbols for “sustained” character contours (rise, high monotone, descent) and the major-stressed syllables initiating them
6x, (x, )x
combined symbols for “end-falling” character contours (rise–fall, monotone-fall, descent-fall) and the major-stressed syllables initiating them
4-x
combined symbol for the stylized falling character contour and the major-stressed syllable initiating it
3x
beginning of the appended contour
]x
minor-stressed syllable
[x
major-stressed syllable when we do not wish to indicate the character contour that it initiates
No symbol (before x)
zero-stressed syllable
(ii)
Pause prosodemes
No symbol
pause lack of pause xi
List of Intonational Transcription Symbols
xii
(iii)
Upstep prosodemes
8
x
(between contours of the front-falling type, contours of the end-falling type, or between identical sustained contours)
the peak of the second contour is higher in pitch than the peak of the first
8
x
(in other cases)
the affected contour appears with a peak height that it would have at the beginning of an utterance
(between contours of the front-falling type, contours of the end-falling type, or between identical sustained contours)
the peak of the second contour is at the same pitch height as the peak of the first
=
x
No symbol (before x)
the contour is downdrifted
Other Symbols and Abbreviations AC ACC AP App BA BD BM C CB CD CONJ CSR DA DEM dep dom DTE EB F f Ft FW FWDA H H8 H9 HRS IMP INF IP L L8 L9 LF LFE MB
association convention accusative accent phrase apposition beat addition beat deletion beat movement comment complementary block clash deletion conjunction compound stress rule deaccentuation rule demonstrative dependent dominant designated terminal element equivalent block focus final (CB/EB) foot function word function word deaccentuation high raised H lowered H highest-ranking sentence imperative infinitive intonation(al) phrase low raised L lowered L logical form lexical foot erasing minor tonosyntactic block xiii
xiv
Other Symbols and Abbreviations
nf non-RV NP NRRC NQIO NSR =obl⇒ OCP =opt⇒ P1 PC PCSR PERF PF pl Pv Q QIO QuoUtt RRC RSSA RV s sg Sn S n′ S n″ ST σ T T UFE Utt V VP W + $ *
non-final (CB/EB) not rhythmically variable noun phrase non-restrictive relative clause non-quoting inorganic block nuclear stress rule ‘obligatorily becomes’ obligatory contour principle ‘optionally becomes’ underlying sentence-level phonological representation phrasal compound phrasal compound stress rule verbal prefix with a perfective meaning phonetic form plural postverb(al) quantifier quoting inorganic block quotation-carrying utterance restrictive relative clause rhythmical secondary stress assignment rhythmically variable syllable string singular syntactic surface structure stressed syntactic surface structure fully intonated syntactic surface structure syntactic type syllable topic unspecified tone ultimate foot erasing utterance verb verb phrase lexical word or non-clitic function word stylizer contour end (after H or L) the tone so marked is to be associated with a major-stressed syllable
Other Symbols and Abbreviations
* *
(over a vowel letter, in ordinary orthography) pitch accent mark (before an example) unacceptable
xv
Hungarian Letter-to-Sound Correspondences
Vowel letter a á e é i í o ó ö o u ú ü u
IPA symbol ! a: ε e: i i: o o: ø ø: u u: y y:
Consonant letter p b t d ty gy k g f v sz z s zs h c
IPA symbol p b t d c @ k g f v s z ∫ # h ts xvi
Hungarian Letter-to-Sound Correspondences
cs dzs m n ny l r j/ly
t∫ d# m n % l r j
xvii
This page intentionally left blank
1 Introduction
1.1
Aims and scope
The aim of this book is two-fold. Its primary aim is to provide the international community of phonologists with a comprehensive description of the intonation and stress system of Hungarian, which is reliable, adequate to the complexity of the subject and clear enough even for readers who have no knowledge of Hungarian. Secondly, but no less important, the book is meant to be a contribution to current theoretical thinking on intonation and stress in generative linguistics. Intonation will be defined as those pitch variations that result in discrete, categorically different vocal effects called melodic prosodemes and upstep prosodemes, functioning in utterances. Though the emphasis in the book will be on intonation, the stress system will also be discussed at length since it is closely related to intonation. Hungarian (Magyar) is a Uralic (or more precisely Finno-Ugric) language spoken by over 13 million people as their native language in Central Europe. It is unlike other European languages, but some features of its stress system and especially of its intonation show remarkable similarity to their counterparts in other languages. Intonation is part of prosody. Hungarian prosody is not much discussed in the international phonological literature. The precious little that has been published in English includes a few overviews: Varga (1983, 1985), Kornai and Kálmán (1989), Hetzron (1992), Fónagy (1989, 1998), sections in Bolinger (1989) and Ladd (1996); and some studies dealing with specific aspects of Hungarian prosody: Kiefer (1967), Kerek (1971), Vogel and Kenesei (1987, 1990), Kenesei and Vogel (1989), Kenesei (1998), Varga (1995, 1998). The segmental aspects of Hungarian phonology have fared better – for instance, the phenomenon of Hungarian vowel 1
2
Intonation and Stress: Evidence from Hungarian
harmony has generated a large number of studies in English – and a comprehensive Hungarian phonology (of vowels, consonants and syllables) appeared in English at the turn of the millennium (Siptár and Törkenczy 2000). By offering a phonology of the most important suprasegmental features of Hungarian, the present book is intended to complement the largely segmental phonological analyses of Hungarian that are available to the English-speaking public. At the same time, the analysis is embedded in the general framework of intonational phonology, and by trying to give its own answers to current theoretical questions, it goes far beyond the concerns of a language-specific description. Cruttenden (1997: 178) enumerates six areas of intonational research (first identified in the 1986 edition of his book) in which considerable progress has to be made if a theory of intonation is to be established. These are: (i) the role of syntax in intonation-groupings and nucleus placement, “with readjustment rules which take account of the discourse environment and which also allow for an element of speaker choice”; (ii) the set of tones (intonation contours) that constitute the intonational lexicon of an individual language (both in terms of contours and in terms of level sequences); (iii) the semantics (abstract meanings) of the set of tones within the intonational lexicon; (iv) the pragmatics (local meanings) of the set of tones in the intonational lexicon, resulting from “the interaction between the abstract meanings of the tones and other levels of meaning (lexical, grammatical, gestural)”; (v) “the realisation rules involved in mapping the tones from the intonational lexicon onto varying stretches of segments which have pre-assigned stresses”; (vi) a comparative study of the preceding areas “to refine our intonational typology and our knowledge of universals”. Cruttenden (1997) adds that work on intonation in the past decade has concentrated almost entirely on a single area, viz. (v) above, where the emphasis has been on the computer implementation of some intonation models and on the phonetic realization of underlying tones. Work in the other areas has unfortunately been lagging behind. “The hope is that this lack of progress will be rectified in the next decade” (1997: 179).
Introduction
3
The present book is a step in this direction. By concentrating on Hungarian intonation, hopefully it makes a significant contribution to all of these areas (with special reference to the first four but also of relevance to the others); moreover, it also deals with the area of the stress system, which is interrelated with intonation. In doing so, it is hoped that the book will contribute to the eventual formation of a theory of intonation. The book is in agreement with many points that have already been made concerning the intonation and stress of other languages, and thus strengthens the consensus about many aspects of a future theory. At the same time, however, some of the claims that the book makes may be considered surprising and provocative because they seem to deviate from current theoretical expectations, though they follow from the facts of Hungarian. The most radical claims of this kind include the following: (a) there is a right-to-left dependency of intonation contours within the Hungarian sentence; (b) Hungarian stress is not fully relational and thus in Hungarian there is no need for a metrical grid as an analytic device. It would be necessary to survey a large number of other languages to see if similar claims can be made outside Hungarian, but such an investigation will be left for future research. The book consists of two parts and nine chapters. Chapter 1 introduces the book itself and contains important background information. The next four chapters constitute Part I and they analyse intonation. Chapter 2 is an introduction to prosody, paralanguage and intonation. This is necessary because many scholars in Eastern Europe do not clearly separate intonation from other aspects of prosody or from paralanguage, and even Western scholars seem to disagree on the exact nature of this separation. Therefore a clarification of this book’s definitions of these concepts is in order. Chapter 3 is a taxonomic analysis of the prosodemes of Standard Hungarian, based on surface contrasts. Twelve meaningful intonation contours (melodic prosodemes) will be distinguished and their forms and meanings presented in an intonational lexicon. The melodic analysis in Chapter 3 is contour-based (configurational) rather than level-based. The chapter will also present the stress-, pause- and upstep prosodemes of Hungarian, and the structure of the Hungarian intonation phrase. Chapter 4 revisits the system of intonation contours and analyses their genesis and form in the theoretical framework of autosegmental phonology. Chapter 5 examines the relation of intonation to syntax and attempts to establish rules for the melodic segmentation of the Hungarian utterance. The next three chapters concentrate on stress, and form Part II of the book. Chapter 6 shows why it is necessary to distinguish no less and
4
Intonation and Stress: Evidence from Hungarian
no more than three phonological stress degrees in Hungarian, and then deals with word stress, and the stressing of phrases and sentences. Chapter 7 demonstrates that the different strengths (intensities) of the major stresses are irrelevant in Hungarian phonological operations such as rhythmical variation in the stressing of phrasal compounds. Chapter 8 turns to the controversial question of rhythmical secondary stresses and provides evidence for our contention that – contrary to some views – minor stresses, including rhythmical secondary stresses, are phonologically relevant in Hungarian: the rule of Contour Insertion is sensitive to them. Contour Insertion is an intonational rule which seeks a minorstressed syllable, and so it also demonstrates the interrelatedness of stress and intonation. Finally, Chapter 9 provides a summary of the main points of the book.
1.2
Theoretical assumptions
According to Goldsmith (1982: 422) “[i]ntonation is the Golden Triangle of Linguistics; it rests at the spot where syntax, phonology, and semantics uneasily come together”. Although there have already been sporadic attempts to account for certain aspects of Hungarian prosody in a nonderivational, constraint-based framework,1 such approaches are as yet unable to offer a unitary explanation of the whole complexity of intonation and stress, with all its phonetic-phonological, syntactic and semantic-pragmatic ramifications. The aims of the present book are therefore better served by using a derivational, rule-based framework. The account of Hungarian stress and intonation offered in this book is based on three theoretical assumptions. 1.2.1
Melodic prosodemes as meaningful wholes
The first assumption, derived from the ideas to be found in Gussenhoven (1983a, 1985), and to some extent in Bolinger (1986, 1989), is that melodic prosodemes (intonation contours) are like lexical items because (a) they can be simple and derived; (b) they are meaningful wholes; and (c) their autosegmental representations, together with some representation of their meanings, are stored in an “intonational lexicon”, from which they can be taken and assigned to syllable strings within utterances. Most melodic prosodemes, notably the ones that we shall call character contours (see section 3.2), are associated with major-stressed syllables; that is they start on a major-stressed syllable and continue until the next major-stressed syllable or, if there is no other major-stressed syllable, until the end of the intonational phrase. Major-stressed syllables are doubly
Introduction
5
prominent. First, they have some extra intensity in relation to the intensity of unstressed syllables, and so they start feet. (A foot lasts from one stressed syllable until the next one, or – in contexts lacking another stressed syllable – until the end of the intonation phrase.) Secondly, major-stressed syllables receive special prominence from the melody which they initiate: they are pitch-accented (or just “accented”, for short). By contrast, minor stresses do not initiate character contours, only feet. 1.2.2
Syntax–phonology mapping
The second assumption is that the prosodic shaping of an utterance is superimposed upon, and to some extent determined by, the syntactic surface structure of the sentence and it evolves in several steps during the syntax–phonology mapping. This assumption regards intonation as being sentence-based rather than discourse-based. This is not to deny that there may be significant connections between intonation and discourse (cf. Brazil, Coulthard and Johns 1980; Brazil 1985, 1995). But the connections between sentence structure and intonation are also quite numerous and have not yet been fully explored, let alone systematically described for Hungarian. In this respect the present book’s view on syntax–phonology mapping is related to that of Selkirk (1984), and assumes the essential validity of the T-model, see (1)a, in which PF (phonetic form) receives information directly from S-structure (roughly: the syntactic surface structure), and the LF (logical form) does not directly affect PF. 2 The LF is crucial for the meaning of a sentence but is not identical with it.
(1)a.
D-structure
S-structure Meaning
LF
PF
However, there are two points at which we deviate from Selkirk’s approach to the syntax–phonology mapping. The first point is that the PF – more precisely, the intonation within the PF – is only partly determined by syntactic structure and is partly independent of it. Selkirk also recognizes this, but dismisses the syntax-independent part as the “expressiveness component” of intonation, which she declares may be largely ignored, as opposed to the “information structure or focus structure component” of intonation, to which she pays considerable attention.
6
Intonation and Stress: Evidence from Hungarian
Selkirk even relegates the rising intonation of English yes–no questions to the expressiveness component (Selkirk 1984: 199, 435). This is possible if one restricts one’s attention to English, where declarative sentences and the corresponding yes–no interrogatives differ syntactically with the result that their intonational signalling is merely attitudinal. But it cannot be accepted for Hungarian (or for Russian, for instance), where declaratives and their yes–no question counterparts are in most cases syntactically identical and it is only their fixed intonation that distinguishes them. Therefore, we assume that that part of intonation which expresses attitudes and communicative sentence types (cf. sections 2.3.2 and 2.3.3) without a “trigger” in the S-structure is the special contribution of intonation to the full meaning of the utterance. To indicate that this syntax-independent contribution also has to be taken into consideration for a full semantic interpretation of the utterance, we modify the T-model by the introduction of a dotted line connecting PF and meaning, as shown in (1)b.
(1)b.
D-structure
S-structure Meaning
LF
PF
The other difference between Selkirk’s approach and ours concerns the order of the steps within the syntax–phonology mapping. Selkirk (1984: 203) proposes “a theory of the intonation–stress relation that puts intonation first”, whereas we suppose that the assignment of stresses precedes the assignment of intonation contours. That is to say, the syntactic surface structure, in which the informationally “new” constituents are marked, is first submitted to stress fixing, where the degree of stress brought into the construction by the inserted words from lexical phonology will change or survive. Then it is submitted to melodic segmentation, in which the intonation contours and intonational phrase boundaries are assigned. In both stages the syntactic bracketings and labels are supposed to be present, and they will be deleted only at the end of the syntax–phonology mapping. That is to say, when the utterance realizes a sentence, the input to the prosodic shaping is the syntactic surface structure (Sn) of the sentence. This is enriched by “a fully elaborated focus structure” (i.e. the marking
Introduction
7
of informationally “new” constituents of the sentence; cf. Selkirk 1984: 206). The syntactic surface structure first undergoes stress fixing. In the stage of stress fixing certain word-initial syllables will retain their major stress, that is their pitch accent (*), inherited from the lexical phonology (see sections 6.5 and 6.6). The pitch accent mark signals that the syllable bearing it will become the starting point of a particular intonation contour (character contour) in the stage of the melodic segmentation. Simultaneously, other word-initial syllables will lose the major stress that they have brought with themselves from the lexical stratum of the phonology, and thus become deaccentuated, that is reduced to minor stress. This means that they lose their pitch accent (*) but retain their foot-initiating capacity. The opposite can also happen: a clitic word, which arrives with zero stress, becomes major-stressed when it is put in the focus position of the sentence (for this position, see section 1.2.3). Stress fixing results in a so-called stressed syntactic surface structure (Sn′), see (2).
(2)
Nem léphetnek egyszerre, mert összedol not step-can-3pl simultaneously because collapses ‘They can’t walk in step because the bridge collapses.’
3
a híd. the bridge
S VP V′ V
Sn′ :
* Nem
léphetnek V
XP
* egyszerre, Pv
S
mert
* összedol
* 4 a híd. Pv
C
Then, in the stage of melodic segmentation, the so-called minor tonosyntactic blocks of the sentence will first be demarcated. By minor tonosyntactic
8
Intonation and Stress: Evidence from Hungarian
blocks (or minor blocks: MBs) I mean those major-stressed syntactic constituents that are the immediate or non-immediate constituents of sentences (see section 5.4). For example, the sentence in (2) is divided into two MBs, and each of these is, in turn, divided into two MBs again, see (3)a. The last (rightmost) MB in a matrix tonosyntactic block is always the dominant MB, while its sisters to the left are dependent MBs.
(3)a.
* Nem
léphetnek
{MBdep}
* egyszerre, {MBdom}
* összedol
mert
{MBdep}
{MBdep}
* a híd. {MBdom}
{MBdom} {MBdom} = Sentence
Then, in the second step of the melodic segmentation, the last majorstressed syllable of the highest dominant MB of the sentence will receive an intonation contour, which is the speaker’s free choice. This is a meaningful choice; the intonation contour chosen is required by the communicative type (direct illocution) and special attitude that the speaker wants the sentence to have. This last intonation contour phonologically constrains the melodic possibilities before it within the sentence. The process takes place in such a way that in each series of sister MBs the last intonation contour of the dominant (last) MB determines what particular intonation contour will be assigned to the last major-stressed syllable within the sister MB(s), or else it leaves that major-stressed syllable temporarily without an intonation contour. The process starts at the top and is repeated at each lower level until the lowest MBs are reached. In the example we have just seen, for instance, the last major stress within the highest dominant MB (i.e. within the sentence) will be given a fully falling intonation contour because the speaker wishes to use the sentence as a statement and this requires a full fall. (A full fall is an intonation contour that falls down to the bottom pitch of the speaker’s voice, cf. section 3.2.2). This is the only melody which the speaker is entirely free to choose within this sentence. The other melodies will all be constrained, either directly or indirectly, by this choice. When this closing melody has been assigned, it may assign a rising intonation contour to the last major stress in its sister MB, or it may leave that
Introduction
9
syllable without a melody for the time being. Our example shows the case where the sister MB receives a rising intonation, cf. (3)b. Since the full fall, symbolized as [ 1], and the rise, symbolized as [2], can only occur in the final portion of an intonation unit, the assignment of such melodies also, automatically, brings in its wake the assignment of an intonational phrase boundary, symbolized as [ |]. (For these and other notational symbols see the List of Intonational Transcription Symbols on p. xi.)
* (3)b. | Nem
léphetnek ´egyszerre, | mert
* összedol
a
`híd. |
In the next step of melodic segmentation, all major-stressed syllables that have remained without an intonation contour will automatically receive a half falling contour, symbolized as [[]), that being the default melody initiated by major-stressed syllables. (A half fall is a falling contour which is similar to a full fall but which does not reach the bottom of the speaker’s voice, cf. section 3.2.2). Our melody rules will have to define what particular intonation contours can or must be assigned to the major-stressed syllables of the dependent blocks and under what circumstances. When the intonation contours have been assigned by the melody rules, pauses may be inserted. The result of melodic segmentation is a fully intonated syntactic surface structure (Sn″), which at the same time is the underlying sentence-level phonological representation (P1); that is S n″ = P 1.
(3)c. Sn″ : | [Nem léphetnek ´egyszerre, | mert [összedol a `híd. | Nem
e
err
sz léphetnek egy
mert
ös szedol
a
h
íd.
The upper half of (3)c exemplifies the use of the intonational transcription system adopted in the book: the symbols mark major stress and melody simultanously. The lower half shows the schematic pitch curve by making the written line of the text move up or down according to the pitch movement of the utterance, and the solid horizontal line represents the bottom of the speaker’s normal voice range. The intonational transcription symbols will ultimately make such diagrams superfluous but, at this stage, while we are introducing the symbols, the diagrams are also necessary.
10
Intonation and Stress: Evidence from Hungarian
1.2.3
The syntactic model chosen
Our third assumption is that for the representation of the syntactic structure of Hungarian, a non-configurational language, the syntactic model put forward by É. Kiss (1981, 1987, 1992) is the most appropriate one. In this model the Hungarian sentence consists of a topic and a comment (defined in purely structural terms). The T or topic position is occupied by one or more or nil T constituents, in any order. The comment (C), that is the highest-ranking VP, has four syntactic positions: (a) the Q or quantifier position, occupied by one or more or nil Q constituents, whose order is subject to scope restrictions; (b) the F or focus position, filled by one or nil F constituent; (c) the V or verb position, filled by the V constituent; and (d) the Pv or postverbal position, occupied by one or more or nil Pv constituents, in any order. The V constituent is an inflected verb, the others (i.e. those filling the T, Q, F or Pv positions) are nonverb sentence constituents. The Q constituent is a universally or near-universally quantified expression (e.g. mindenki ‘everybody’, teljesen ‘completely’, sokan ‘many’, nagyon ‘very much’). The T and Pv constituents can be subordinate clauses, too. The appearance of a subclause in Pv position divides the comment into a presubclausal part and the subclause (and occasionally a post-subclausal part as well). All the syntactic positions are illustrated in (4). As can be seen from the illustration, the XPs are generated behind the verb and then move into their respective surface positions (T, Q, F), or remain behind the verb (in Pv). (4) is a neutral (non-emphatic) sentence: it has “broad focus” semantically, that is the whole sentence or at least the whole comment in it conveys new information. Its syntactic F position is filled by a noun with no determiner (kávét), and each stressable constituent (apart from the verb after the filled F position) is major-stressed. The noun without a determiner is a kind of “reduced complement” (see section 6.6.1). A reduced complement (if there is one), must go into the F position in a neutral sentence.
(4) János mindig kávét iszik John always coffee-ACC drinks ‘John always drinks coffee in the snackbar.’ * * * * János mindig kávét iszik a büfében.
a the
büfében. snackbar-in
Introduction
11
S XP
VP XP
VP V′
XP Jánosi mindigj
V
XP
XP
XP
XP
iszik
a büfében.
0i
0j
0k
kávétk
Q
F
T
V
Pv C
Sentence If the reduced complement in the syntactic F position is the only major-stressed constituent within the comment, (cf. kávét in (5)), then this constituent alone is semantically focused: semantically the sentence has “narrow focus” on the reduced complement:
(5)
* János mindig F[kávét] iszik a büfében.
When the XP in the syntactic F position is other than a reduced complement (cf. a büfében in (6)), the sentence has narrow semantic focus on this XP, and the reduced complement (if there is one) must occur elsewhere than the F position:
(6)
János mindig
F[a
* büfében] iszik kávét.
When the constituent in F position is a predicative noun phrase or adjective phrase, the 3rd person present tense copulative verb (van ‘is’, or vannak ‘are, 3pl’) does not appear on the surface; compare example (7).
12
Intonation and Stress: Evidence from Hungarian
(7)
János diák Debrecenben John student Debrecen-in ‘John is a student in Debrecen.’ * * János diák
* Debrecenben
S XP
VP V′
XP
Jánosi
V
XP
XP
XP
diákj
---
Debrecenben
0i
0j
F
V
T
Pv C Sentence
As has been seen, syntactic focus in Hungarian arises as a result of movement rather than as a result of simple pitch accent placement, that is Hungarian has “ex situ” focus (Kenesei 1998). Syntactic focus and semantic focus do not necessarily coincide: the Hungarian sentence may have broad semantic focus, with several XPs and the verb being “new”, while its syntactic focus always contains only one XP.
1.3
Sources
The statements about Hungarian intonation and stress in this book derive from various sources. The primary source is a large body of data which I have collected over the years by observing my own speech and the speech of the people around me. This kind of speech belongs to Standard (i.e. non-dialectal) Hungarian. 5 I have also used a corpus consisting of two recordings. The first one is a 50-minute spoken narrative, which was formerly available on an LP
Introduction
13
gramophone record (Hungaroton SLPX 13855) containing the text of the Hungarian translation (by Marcell Benedek) of Dickens’ A Christmas Carol, read by a famous Hungarian actress, Mari Törocsik. The other recording is of a 20-minute spontaneous debate recorded live from the closed-circuit TV coverage of a local council meeting in a new suburban housing estate called Gazdagrét, in Budapest. This recording was analysed from a number of aspects as part of a project of the Linguistics Institute of the Hungarian Academy of Sciences.6 The present author made an auditory analysis and a full transcription of the intonation of both recordings. The transcription of the Gazdagrét material has been checked and confirmed by a number of experts working on Hungarian intonation and the pauses in this material have also been instrumentally measured. These two recordings both represent Standard Hungarian. In addition, I submitted to instrumental analysis my own renderings of each Hungarian prosodeme (at the Linguistics Institute of the Hungarian Academy of Sciences and also at the Phonetics Department of Eötvös Loránd University), and relied on the acoustic records so obtained when describing the phonetic realizations of prosodemes. I have also carried out an experiment, involving 105 participants, to verify my hypothesis that people may use certain minor-stressed syllables as hosts for Contour Insertion. This experiment will be described in Chapter 8. Last but not least, I have benefited greatly from past work on stress and intonation, particularly from Molnár (1954), Deme (1962), Fónagy and Magdics (1967) and Hetzron (1980), all of which have been invaluable sources of data and inspiration. In establishing my theoretical stance I have been strongly influenced by É. Kiss (1981, 1987, 1992), Gussenhoven (1983a, 1985, 1991), Bolinger (1986, 1989), Selkirk (1984), Hayes (1984, 1990, 1995) and Vogel (1987). The material in this book has been developed gradually in many of my earlier writings, especially Varga (1981, 1989, 1993, 1994, 1995, 1998, 2000).
1.4
Past work on Hungarian stress and intonation
Studies on Hungarian intonation and stress have a long and rich descriptive tradition. Unfortunately, most of the works in this tradition are inaccessible to an international readership because they have been published only in Hungarian. The present section is an attempt to summarize briefly the most important developments in the field over the last two centuries, with an emphasis on work that has been influential in shaping phonological thinking on these matters.
14
Intonation and Stress: Evidence from Hungarian
Prosodic research in the nineteenth century centred on the stressing of Hungarian sentences and its impact on word order. As early as 1838 János Fogarasi formulated what became known as Fogarasi’s Law: “the emphasized word is followed by the verb, in finite form; the rest of the words may stand before or after, and may follow in any order”.7 Sámuel Brassai (1888, 1889) was also of the opinion that the major stress is the key to word order, and he claimed that there is only one major stress in a Hungarian sentence, the one on the constituent at the beginning of the comment (see section 1.2.3), and that the constituents before and after this may only receive weaker stresses or no stresses. Others (e.g. Simonyi 1880, Joannovics 1889, Balassa 1887–90, Kicska 1891), however, maintained that the stresses on other constituents in the sentence may be just as strong as the one at the beginning of the comment. To some extent this controversy is still with us today. Katalin É. Kiss, who worked out the first generative syntactic model for Hungarian (É. Kiss 1981), revived Brassai’s view of the Hungarian sentence always having one single major stress on the comment-initial element, and attempted to compute the stress degrees of words in Hungarian syntactic constructions by applying to them a Hungarian version of the Nuclear Stress Rule (É. Kiss 1987–88). This engendered considerable criticism (Varga 1986, 1987–88; Kálmán et al. 1989; Kálmán and Nádasdy 1994). A reasonably comprehensive description of the stressing of Hungarian phrases and sentences was given by László Kálmán and Ádám Nádasdy (1994). There is a growing consensus today that there are special emphatic sentences with a single major stress, but there are also sentences with a “flat” prosody, that is with several equally strong major stresses (Fónagy 1998). Studies of speech melodies on a larger scale began only at the start of the twentieth century. Since then, the overwhelming majority of intonational works have been conceived in the “configurational” framework, that is a framework which treats speech melodies as configurations or contours rather than sequences of phonemic levels. When these studies refer to levels, they are used only as descriptive devices. The first significant development in intonational research at the beginning of the twentieth century was associated with the name of Vilmos Tolnai, who introduced a kind of quasi-musical notation for intonation. Although Tolnai distinguished four pitch levels, they did not represent absolute pitch values (Tolnai 1915). Tolnai’s system was further developed by Bálint Csury, who used five pitch levels and produced an excellent description of the emotional and non-emotional pitch patterns of the dialect of Szamoshát, a region in north-eastern
Introduction
15
Hungary (Csury 1925). Later Csury (1935, 1939) turned to the investigation of Standard (i.e. non-dialectal) Hungarian intonation, and claimed that the intonation of Hungarian sentence types appears in a few permanent patterns. Gyula Laziczius (1944: 186) also emphasized the importance of these patterns in any investigation of intonation. A significant step towards a more abstract approach to intonation was made by Imre Molnár (1954). Instead of the inductive method of his predecessors, Molnár took a deductive approach: he was interested in the forms and meanings of the speech melodies that are possible if three pitch levels are postulated. László Deme (1962) systematized many of the intonation patterns of Standard Hungarian, and presented their non-emotional and emotional variants in the various grammatical sentence types. He also emphasized the importance of general pitch heights (registers). Initially Deme identified the unit of intonation with the foot (1962), but subsequently he realized that the units of intonation should be able to contain sequences of feet (1979). Deme analysed intonation within the broader context of “sentence-phonetic” devices but warned that the analysis might have to go beyond the sentence (1970). This latter view was fully adopted by Imre Wacha (1985), who described intonation as a subset of “text-phonetic” devices. László Elekfi (1964, 1968) investigated how the topic–comment division of the sentence was mirrored in the intonation of declarative sentences. Iván Fónagy and Klára Magdics’s (1967) book presented the results of the greatest empirical investigation of Standard Hungarian intonation ever carried out, and is an inexhaustible source of information about Hungarian intonation even today, although it has been criticized for its musical notation. From the mid-1970s onwards, the phonological aspects of intonational research, which had already appeared in Molnár’s and Deme’s work, grew stronger. Endre Kozma (1974) surveyed the speech melodies of Hungarian using the theory of the Soviet researcher Bryzgunova. The descriptive apparatus of Endre Kozma and Tamás Szende (1981) showed a belated influence of the structuralist approach to intonation treatment prevalent in the USA in the 1950s. In its heyday, structuralism was not considered acceptable in Hungarian linguistics, due to the hostile political and ideological climate in Eastern Europe at that time. Another version of the structuralist approach to intonation, showing the effects of SPE-type linear phonology, appeared in Robert Hetzron (1980). The first autosegmental treatment of Hungarian intonation was produced by
16
Intonation and Stress: Evidence from Hungarian
László Kálmán and András Kornai (1985). László Varga (1981, 1983, 1985) has not restricted his attention to speech melodies but has tried to characterize the whole complexity of Hungarian intonation and related fields in terms of phonological oppositions, by describing prosodemic systems (of melodic, stress, pause, and upstep prosodemes), identifying the unit of intonation, and introducing notational symbols by means of which the intonation of large stretches of speech could be transcribed without supplying separate diagrammatic representations. Varga also examines the relation between syntactic structure and intonation, and suggests an alternative autosegmental analysis of Hungarian intonation (Varga 1993, 1994). He describes the stylized falling contour of Hungarian (1995), analyses rhythmical variation in certain Hungarian compounds (1998), and identifies several kinds of minor stress in Hungarian (2000). Most recently, minor stresses and feet in Hungarian words have been the topic of Zoltán Gráf ’s (2001) constraint-based research. László Hunyadi (1999) deals with the question of how scope relations are reflected by prosodic structure. T. Eniko Németh (1996) examines the role of intonation in breaking up spoken discourse into utterances. This, briefly, summarizes the evolution of the phonological approach to intonation and prosody in Hungary. Without these works, the present book could never have been written.
Part I Intonation
This page intentionally left blank
2 Intonation, Paralanguage, Prosody
2.1
Introduction
In this chapter we first clarify, in section 2.2, some basic concepts that are necessary for defining intonation, and accommodate intonation within the broader framework of prosodic devices. The term prosody will be used in its broadest sense, as a loose synonym for suprasegmental or non-segmental phonation. Intonation is defined as those pitch variations that are the manifestations of melodic prosodemes and upstep prosodemes, functioning in utterances. Intonation is closely related to stress prosodemes and pause prosodemes, but has to be separated from paralanguage and also from the prosodic limits within which particular speakers use it. Section 2.3 identifies four types of intonational function – attitudinal, cognitive, thematic and communicative sentence type distinctions – and demonstrates the division of labour between syntax and intonation in the expression of certain meanings.
2.2 2.2.1
Prosodic devices and intonation Intonation as part of prosody
Intonation can be identified with a subset of prosodic (or suprasegmental) devices. These are the results of stereotypical variations in pitch, loudness, duration and quality, superimposed upon different-sized sections of the sound string (Cruttenden 1997: 172–7). 1 The prosodic devices include pitch patterns (i.e. recurring configurations of consecutive pitch heights), pitch range (i.e. the distance between the highest and lowest pitch heights), pitch register (i.e. the general pitch level of a section of speech), stress, volume, tempo, voice quality variation and pause. 2 Pitch patterns, pitch range and pitch register constitute various kinds of pitch variation. 19
20
Intonation and Stress: Evidence from Hungarian
The stereotyped variations in some prosodic devices produce discrete, categorically different vocal effects; these will be called prosodemes. Different approaches identify intonation with different subsets of prosodic devices. Some scholars adopt a broad definition, according to which intonation comprises all the prosodic devices taken together (including even voice quality variation), for example Bolla (1980), Cheremisina (1982). Most scholars, however, define intonation narrowly and reduce it to the use of pitch variation (i.e. pitch patterns, pitch range, pitch register; with pitch patterns playing the central role), with the proviso that the use of pitch is relevant not at word level but at the level of utterances (spoken sentences), for example Jones (1967: 275), O’Connor (1973: 190), Tench (1996: 2–6). When pitch patterns are used for distinguishing utterances, they can be called intonation contours (or speech melodies).3 But even the narrow definitions agree that pitch variation interrelates with other prosodic devices, especially stress, see Daneš (1960). We shall adopt the narrow definition of intonation and reduce it further to those variations in pitch that result in discrete, categorically different vocal effects, performing certain functions in utterances (see section 2.3). The central component of intonation is the system of melodic prosodemes. These are the types of recurring pitch patterns (intonation contours, speech melodies) which speakers realize in their utterances and which enter into paradigmatic contrasts with one another as wholes. In addition to the melodic prosodemes, our definition of intonation also includes the system of upstep prosodemes. These are variations in the height of the peak of an intonation contour in relation to the height of the peak of the preceding intonation contour (it can be higher or lower than, or the same height as, the preceding peak). These two kinds of prosodemes constitute the subsystems of intonation. However, intonation works together with, and is closely related to, two other kinds of prosodemes: stress and pause. Stress prosodemes are the significant degrees of syllabic prominence, achieved by the presence or lack of extra intensity (“extra” in relation to the intensity of unstressed syllables), and, when extra intensity is present, by the presence or lack of intonation contour initiation. A complete lack of intensity (loudness) in the course of speech appears as pause, which – together with the absence of pause – belongs to the set of pause prosodemes. In our approach, the melodic and upstep prosodemes constitute intonation proper, with stress and pause prosodemes closely tied to them, and all four kinds of prosodemes represent discrete, categorical variation. Those prosodic devices that are produced by continuous variation rather than by discrete, categorical variation (viz. variations in pitch
Intonation, Paralanguage, Prosody
21
range, pitch register, volume, tempo, voice quality, intensity of major stresses), however important they may be, are not prosodemes and will be regarded as belonging to paralanguage. Being paralinguistic, they will not be dealt with in this book.4
2.2.2
Prosodic limits
Each speaker has certain individual and language-specific limits within which (s)he can use the prosodic devices (including intonation) of the language. These limits define the boundaries within which a speaker’s prosodic solutions can vary, and are comprised of the speaker’s own average pitch register, pitch range, volume, tempo and voice quality. These are related to the speaker’s anatomical, psychological and social characteristics and together they constitute the speaker’s individual prosodic limits. These limits may provide information to the listener about the speaker’s sex, age, physical build, personality, articulatory disorders and their causes and – in the case of a known speaker – they may even identify the speaker when he or she is not in view. This information is non-linguistic; it is supplementary to intonational information – especially to the attitudinal information conveyed by intonation (see section 2.3.2) – but is not identical to it. There are also language-specific prosodic limits. It has been found that Hungarian speakers – most notably female Hungarian speakers – speak in a considerably lower average pitch register than Russians or Germans (Vértes 1980: 175–6). The average pitch range of English intonation is wider than that of Hungarian intonation. The range of British female speakers in conversation is a musical tenth, while that of male speakers can be even greater (Jones 1967: 276). Most Hungarian intonation contours, however, have a much narrower range than a musical tenth (Fónagy and Magdics 1967). Such facts are important to the interaction between speakers who have different native languages and also in foreign language teaching, but are insignificant within the language community that displays them. Since the language-specific prosodic limits appear to a given linguistic community as the boundaries within which the prosody can carry common, normal psychological content, the information that the individual prosodic limits supply about a speaker may take on different local values in different languages. Consequently, if we want to correctly evaluate foreign speakers’ attitudinal and psychological characteristics by listening to their intonation, we have to take into consideration both their individual and language-specific prosodic limits (Péter 1984: 223).
22
Intonation and Stress: Evidence from Hungarian
2.2.3
Paralinguistic variations
The continuously variable prosodic devices are paralinguistic. We can produce natural and conventional variation in paralinguistic prosody. It is a common experience that – within our individual and languagespecific prosodic limits – our voice automatically becomes higher when we are excited, and the more excited we are, the higher our voice may be. These are automatic reactions to certain stimuli and can be called natural variations in paralinguistic prosody. But it is also possible for speakers to raise their voices deliberately – again, within the available individual and language-specific limits – because the variety so chosen carries a conventional meaning which they wish to convey. For example, a high register is the sign of obligatory politeness in Tamil (Cruttenden 1997: 124). This is a conventional variation in paralinguistic prosody. Although the conventional variations in paralinguistic prosody may stand closer to linguistic communication than the natural ones, they are still paralinguistic and so outside the scope of the present book.
2.3 2.3.1
The functions of intonation The types of intonational functions
Intonational devices perform their functions by interacting with the segmental material (text) of a sentence and by turning the sentence into an utterance. Thus, producing an utterance out of a sentence may itself be regarded as an intonational function (the “constitutive” function of intonation, see Daneš 1960), but at the same time it is a prerequisite of the other, more widely recognized, intonational functions. The functions of intonation are traditionally divided into two groups: “grammatical” and “attitudinal” (Crystal 1969: 272), or “linguistic” and “attitudinal” (Lehiste 1970: 95–6). More recently these have also been referred to as the “linguistic” and “paralinguistic” dimensions of intonation (Tench 1996: 2–3). However, it can be reasonably argued that the expression of attitudes by intonation (i.e. by categorically variable, conventional pitch-related prosodic devices) is also a linguistic matter (Antipova 1979: 23) and so even the “purely attitudinal” functions of intonation can be subsumed under the linguistic functions. Therefore, we shall be using the labels “grammatical” and “attitudinal” for the basic distinction. As we shall see in section 2.3.5, in some cases the two kinds of functions may not be clearly separated from each other and form a cline.
Intonation, Paralanguage, Prosody
2.3.2
23
The attitudinal functions of intonation
By the attitudinal functions of intonation we mean the expression of the speaker’s attitude(s) by intonational means. An attitude is an extremely complex form of behaviour; it is not simply emotion but rather the speaker’s particular mental position during speech, which (s)he adopts in relation to him-/herself, his/her audience, his/her message, the situation, etc. and it may have emotional and/or volitional components, possibly mirroring the social relationships between speaker and audience on the one hand and also the “genre” of the speech event on the other hand. Therefore, attitudes can be seen to have not only psychological but also social and sometimes stylistic aspects. 5 The intonation contours that reflect attitudes are conventional, and so they may differ from language to language in form or meaning. Thus, for example, both English and Hungarian have a falling–rising intonation contour, with a somewhat different meaning, see (1)a and (1)b. The words in (1)a are: beszélgetni ‘talk-INF’, voltam ‘was-1sg’, and the sentence means ‘I have been somewhere to chat a little.’
(1) a. Be
tam. szélgetni vol (Csury 1925 : 31)
b. Some
right. of them are all (O’Connor and Arnold 1961 : 232)
In English, this melody either signals conflict and reservation (it is the “contradiction contour” of Liberman and Sag 1974), or it is associated with “shared knowledge” (i.e. “given” information; see Brazil, Coulthard and Johns 1980; Brazil 1985; Gussenhoven 1983a: 13–15). In Hungarian it may also express conflict (cf. section 3.2.2) but is not typically used as the melody of “given” information, unless it is conflicting at the same time. The fall–rise contour does not exist in Russian, German or Danish (O’Connor 1973: 246). The Hungarian contour for yes–no questions, the characteristic rise–fall (see section 3.2.4), whose form differs from the English rise–fall, is interpreted by English listeners as having a much more commanding attitude than it actually has in Hungarian (Coates 1975).
24
Intonation and Stress: Evidence from Hungarian
The variations of paralinguistic prosody also contribute to passing information about attitudes and in so doing they interact with the conventional melodies. The speaker’s attitude may cause automatic variations in pitch register, pitch range, volume, tempo and voice quality, and these variations appear in similar forms in a wide range of languages. For instance, fear lowers the register, narrows the range, decreases the volume, increases the tempo and brings on a husky voice quality in Hungarian just as in Western European languages (Fónagy and Magdics 1967: 204–5, 263–4). 2.3.3
The grammatical functions of intonation
In its grammatical functions, intonation disambiguates sentences which are ambiguous in terms of their cognitive meaning, thematic meaning or communicative sentence type. Such disambiguation by intonation can be regarded grammatical because similar distinctions can also be brought about by morphological or syntactic means. Cognitive meaning (or conceptual meaning; see Leech 1974: 10–13) is that part of the meaning of a sentence which can be grasped by reference to truth conditions. Distinguishing cognitive meaning by intonational means usually goes hand in hard with the distinguishing of syntactic structure. For instance, (2)a and (2)b differ in cognitive meaning because (2)a implies that Sándor is my friend, while (2)b implies that Sándor is not my friend. But they also differ syntactically: in (2)a Sándorról is an appositive to barátomról, whereas in (2)b Sándorról is a conjoin in an adversative coordination.
(2)
Nem not
a the
barátomról, friend-my-about
Sándorról beszéltek. Alexander-about talked-3pl
a. Nem a barátomról,
Sándorról be szél
tek.
‘It is not my friend, Sándor, that they were talking about (but someone else).’ b. Nem Sán barátomról, a
dorról beszél tek.
‘They were talking not about my friend, but about Sándor.’
Intonation, Paralanguage, Prosody
25
Thematic meaning (Leech 1974: 22–3) is communicated by the way a sentence is organized into informationally given, new or contrasted parts. When this meaning manifests itself in the division of the sentence into topic and comment (see section 1.2.3), it cannot always be separated from cognitive meaning because the topic–comment division may affect the truth value of the sentence (Szabolcsi 1981; Kiefer 1983). In addition, in such cases the intonational differentiation of thematic meaning goes together with syntactic differentiation in Hungarian. In (3)a the NP a piros toll (‘the red pen’) is in topic position, while in (3)b it is in focus position. The 3rd person present tense copula is missing; see section 1.2.3.
(3)
A piros toll the red pen
az the
enyém. mine
a. [Context : Kié a piros toll?] ‘Whose is the red pen?’ A piros
toll az
e nyém.
‘The red pen is MINE.’ b. [Context : Melyik a te tollad?] ‘Which is your pen?’ A
pi ros toll az enyém.
‘The RED pen is mine.’
The term communicative sentence type refers to those finer divisions within the traditional grammatical sentence types which are determined by the speaker’s communicative intention and the context.6 The responses in (4)a and (4)b belong to different communicative sentence types; the one in (4)a is an ordinary question-word question, the one in (4)b is a repetitive question-word question: the speaker finds what (s)he has just heard incredible, or has not heard it properly, and wants his/her partner to repeat it.
26
(4)
Intonation and Stress: Evidence from Hungarian
Mit what-ACC
láttál? saw-2sg
a. [Context: Moziban voltam.] ‘I’ve been to the cinema.’ Mit láttál? ‘What did you see?’ b. [Context: Láttam egy pingvint.] ‘I saw a penguin.’ lát Mit tál? ‘What did you see? = You saw what?’ The variations of paralinguistic prosody (see section 2.2.3) may accompany and confirm not only the attitudinal but also the grammatical functions of intonation. For instance, the less important parenthetical sections of speech have a lower register, lower volume and a more rapid tempo in many languages, including English and Hungarian. On the other hand, a widened pitch range may distinguish an exclamation, (5)a, from a statement, (5)b:
(5)
Felállt a kisbaba up-stood-3sg the small-baby a. Fel kis állt a baba! ‘The baby stood up!’ b. Fel
(Exclamation)
állt a kis ba ba.
‘The baby stood up.’
7
(Statement)
Intonation, Paralanguage, Prosody
27
It may happen that the sentence which forms the basis of the utterance is unambiguous with respect to its cognitive meaning (and syntactic structure), but is still ambiguous with respect to its thematic meaning and communicative sentence type, as in (6)a, (6)b, (6)c and (6)d.
(6)
Megitta PERF-drank-3sg a. Meg itta a
a the
pálinkát. brandy-ACC
pá linkát.
Cognitive meaning: ‘He has drunk the brandy.’ Thematic meaning: the whole sentence is new information. Communicative sentence type: ordinary statement. b. Meg itta a pá linkát. Cognitive meaning: as in (6)a. Thematic meaning: “a pálinkát” is given information. Communicative sentence type: ordinary statement. c.
lin a pá ta it Meg kát? Cognitive meaning: as in (6)a. Thematic meaning: either the whole sentence is new information or “a pálinkát” is given information. Communicative sentence type: ordinary yes–no question.
d. Meg itta a
lin pá
kát?
Cognitive meaning: as in (6)a. Thematic meaning: the whole sentence is given information. Communicative sentence type: echo (surprised) yes–no 8 question.
28
Intonation and Stress: Evidence from Hungarian
The text of the sentence may be unambiguous with respect to both its cognitive and thematic meaning, but the disambiguation of the communicative sentence type may still have to be done by the intonation, as in (7). By putting the verbal prefix meg after the verb itta, we signal that the constituent before the verb is in focus. (For the position of the verbal prefix, see section 6.6.1.)
(7)
A pálinkát itta the brandy-ACC drank-3sg a.
meg. PERF
pá A linkát itta meg. Cognitive meaning: as in (6)a. Thematic meaning: “a pálinkát” is focal information. Communicative sentence type: ordinary statement.
b.
kát A pálin
itta meg?
Cognitive meaning: as in (6)a. Thematic meaning: “a pálinkát” is focal information. Communicative sentence type: ordinary yes–no question. Finally, there also exist sentences that are completely unambiguous (i.e. even their communicative sentence type is signalled by the text of the sentence) and so the intonation has no grammatical function to fulfil. (But it may still convey attitudinal distinctions.) This is the case in (8). The only new word is ugye (approximately: ‘is it so?’), which signals that the sentence it introduces is a biased yes–no question.
(8)
pá Ugye a linkát itta meg? Cognitive meaning: as in (6)a. Thematic meaning: “a pálinkát” is focal information. Communicative sentence type: biased yes–no question (expecting confirmation).
Intonation, Paralanguage, Prosody
29
As will be shown in Chapter 5, intonation is partly determined by syntactic structure, and is partly independent of it. This is why the syntactic structure and the intonational realization both have to be taken into consideration when we want to interpret an utterance from a semantic–pragmatic point of view. The more strongly a certain meaning is expressed by syntactic devices (e.g. by word order), the weaker may be its intonational expression, and vice versa: the stronger the intonational expression, the weaker may be the syntactic signalling (Peshkovskij 1959). This division of labour between syntax and intonation can be exemplified by comparing yes–no questions in Hungarian and English. Ordinary Hungarian yes–no questions are not marked syntactically (they do not differ from statements syntactically) but are marked by a fixed rising–falling intonation, while statements have a falling intonation (cf. sections 3.2.2 and 3.2.4). In English, however, ordinary yes–no questions have a special word order and so their intonational realization is quite flexible. 2.3.4
The primacy of melodic prosodemes
In any case, whether we consider the grammatical or the attitudinal functions of intonation, it is the melodic prosodemes that have the greatest functional significance (cf. Juhász 1968: 224–5, Greenberg 1969: 61, Bryzgunova 1971: 43, fn., Torsujeva 1979: 94). Moreover, speech melodies have an indispensable role in language acquisition and the speech comprehension of small children (Kassai 1979). The central role of melodic prosodemes is also shown by the fact that they are the only kind of prosodemes included in all intonation definitions. 2.3.5
Other groupings of intonational functions
Other ways of classifying intonational functions are also possible. For instance, Roach (1983: 136–7) distinguishes the attitudinal, accentual, grammatical and discoursal functions of intonation but because of the increasing amount of overlap in the last three groups, he eventually brings them together under the heading of the “syntagmatic functions” of intonation, which he contrasts with the “attitudinal functions”. Tench (1996: 16–29) mentions the information-organizing, communicative, syntactic, attitudinal, textual and speech stylistic functions but in effect he only deals with the first four. These roughly correspond to the four functional categories we have distinguished in sections 2.3.2 and 2.3.3. Tench (1996: 2–3) also separates the attitudinal functions as the “paralinguistic dimension” of intonation from all the other functions, which together constitute the “linguistic dimension” of intonation in his system.
30
Intonation and Stress: Evidence from Hungarian
Let us return to the grammatical functions of Hungarian intonation that we presented in section 2.3.3. The assumption that the intonational distinctions of cognitive meaning and thematic meaning are grammatical functions is uncontrovertible. But when intonation distinguishes communicative sentence types, the boundary between the “grammatical” and “attitudinal” functions is not clear and the two kinds of function form a continuum (Crystal 1969: 272).9 There are several facts that lend support to this view. First, it is likely that the communicative sentence types themselves derive from general, ancient attitude types and that they still carry some inherent attitude. This would account for the surprising similarities that we can observe in different languages in the ways certain intonational features regularly accompany certain sentence types, cf. Bolinger (1978, 1989: 26–66). Since the attitudinal functions of intonation are closer to the unintentional signalling of physiological states (Bolinger 1989: 194–214), it seems reasonable to assume that they are older than the grammatical functions. That is to say, in the beginning intonation was natural and attitudinal, but due to the influence of the conventional devices of language, intonation itself became more and more conventional and its grammatical functions gradually emerged (Péter 1961: 134). Secondly, intonation functions attitudinally even in cases where it does not add any extra attitude to (the inherent attitude of) a communicative sentence type, because the unmodified expression of the inherent attitude of a communicative sentence type is also a kind of attitude: the neutral attitude. For instance, we can ask a question eagerly, slyly, sadly, happily, reluctantly, enthusiastically, etc., but we can also ask it “neutrally”, that is without any identifiable extra attitude. However, neutrality may have different pragmatic values in different circumstances. For instance, it may be negatively interpreted in situations where we are expected to show delightful enthusiasm but we choose to be “neutral” instead. Thirdly, intonation is rarely “purely attitudinal”; one such case was shown in example (8). In most cases, however, attitudinal intonation simultaneously also carries out grammatical distinctions of various kinds. Since the boundary between the grammatical and attitudinal functions of intonation is blurred, it is not surprising that some scholars recognize only one group of intonational functions. Halliday (1967) thinks that in language we establish meaning distinctions by either lexical or grammatical means. Since intonation is not lexical, it must be entirely grammatical (even in its so-called attitudinal roles). Others, for example
Intonation, Paralanguage, Prosody
31
O’Connor and Arnold (1961), believe that this extension of the notion of grammar is unjustified and that intonation expresses only attitudes. Similar views were expressed, especially in the early period of his work, by Bolinger. He was of the opinion that the encounters between intonation and grammar are casual, not causal (1957–8: 37), intonation is “around the edge of language” (1964). Later, however, Bolinger acknowledged that intonation has “a highly grammaticized layer” which performs grammatical functions (1972: 137) but these are secondary to the attitudinal functions (1986: 27). To this we can add that the “highly grammaticized layer” of intonation may have different extents in different languages and that in Hungarian this layer is more extensive than in English (cf. Fónagy 1998: 334). In addition to the “all-grammatical” and “all-attitudinal” views on intonational functions there is a third way of asserting the homogeneity of these functions. This approach was started by Brazil in the 1970s and its main claim is that intonation is not a property of the sentence or the sentence-based utterance but of discourse and so its functions are neither grammatical nor attitudinal but “discoursal”: showing information status and regulating conversational behaviour (Brazil, Coulthard and Johns 1980; Brazil 1985, 1995). However, intonation contours do not always correspond to what could be expected on the basis of the discourse functions (cf. Gussenhoven 1983a: 83–5; Cruttenden 1997: 106–8), in addition the discourse functions themselves overlap considerably with subtypes of both the attitudinal and the grammatical functions and so they can also be treated among these (Cruttenden 1997: 89).
2.4
Summary
Since the literature identifies intonation with different subsets of prosodic devices, in this chapter we first overviewed prosodic devices and their discrete variations, the prosodemes. We defined intonation as the pitch-related subset of prosodemes (i.e. as melodic and upstep prosodemes), which, together with stress and pause prosodemes, function within utterances. We separated intonation from paralanguage and also from the prosodic limits within which particular speakers use it, and identified the main types of intonational functions.
3 A Taxonomic Analysis of Hungarian Intonation
3.1
Introduction
The previous chapter gave us a general view of intonation: we considered the functions of intonation and its place in the broader framework of utterance prosody. In the present chapter we turn to Hungarian and carry out a taxonomic analysis of the prosodemes of Hungarian, based on surface contrasts.1 This analysis will provide us with an inventory of the prosodemes which we shall submit to an autosegmental analysis in Chapter 4. The chapter is a synthesis and revision of the relevant parts of Varga (1989, 1993, 1994, 1995). Twelve meaningful intonation contours of Standard Hungarian will be distinguished on the surface (though two of them will coincide in phonetic shape). Each of these contours has a general meaning (the same as Cruttenden’s (1997: 106) “abstract meaning”), which is the common factor in all uses of the contour. The general meaning of each intonation contour will be represented by means of glosses. The result of the analysis will be an intonational lexicon of Hungarian, in which each entry consists of a representation of the phonetic shape of the contour, matched by a representation of its meaning. This intonational lexicon, consisting of intonation contours, will be the analogue of the verbal lexicon, consisting of words.2 The structure of the chapter is as follows. First we discuss the so-called character contours and the appended contour in sections 3.2 and 3.3. We then establish the intonational lexicon of Hungarian in section 3.4. Then, after examining the preparatory contour in section 3.5, we explore the Hungarian intonation phrase and its internal structure in section 3.6. Finally, in section 3.7, we describe downdrift and the upstep prosodemes. 32
A Taxonomic Analysis of Hungarian Intonation
3.2 3.2.1
33
Character contours On character contours and degrees of stress
The twelve meaningful speech melodies we shall distinguish in Hungarian can be divided into eleven character contours and one appended contour. A character contour (or just character for short) is a discrete meaningful speech melody which is able to appear on independent utterances; it has a characteristic shape (involving a high pitch for at least part of the contour) and extra intensity on the first syllable of the phrase carrying the contour. The first syllable is major-stressed because, in addition to its extra intensity, it initiates a character contour, thereby receiving special prominence. Therefore the major-stressed syllable is also said to be pitch-accented (or simply accented), and the terms “major stress” and “accent” can be used synonymously. Any syllable which has extra intensity but does not initiate a character contour will be considered minor-stressed. All other syllables are zero-stressed (or unstressed). Both major- and minor-stressed syllables have extra intensity and consequently start feet. While major-stressed syllables are accented, minorand zero-stressed syllables are unaccented. Accordingly, the system of Hungarian stress prosodemes can be characterized by the binary features of extra intensity and character contour initiation, as shown in Table 3.1. From this definition of the three degrees of stress it follows that the actual differences in the intensity of the major-stressed syllables will be ignored. We shall regard all syllables that have some audible extra intensity and that initiate a character contour as major-stressed, independently of the actual intensity differences that may exist between them. That is to say, we do not distinguish major stresses (accents) according to their strengths. The argument in favour of this analysis is related to the facts of rhythmical variation, and will be discussed in Chapter 7. At this stage we have no more to say about stress; a detailed treatment will follow in Part II, that is in Chapters 6, 7 and 8.
Table 3.1
The system of Hungarian stress prosodemes
Zero stress Minor stress Major stress
Extra intensity
Initiation of character contour
– + +
– – +
34
Intonation and Stress: Evidence from Hungarian
In most cases the shape of the contour alone is sufficient for identifying the contour as a character contour, because its production automatically involves extra intensity on the first syllable of the carrier phrase. For instance, the falling contour, with the characteristic big drop between the first and second syllable of the carrier phrase, is always a character contour. The so-called sustained contours, that is the rise, the high monotone and the descent (see section 3.2.3), however, are capable of appearing without extra intensity on the first syllable of the carrier phrase, and when they do so, they are not character contours but preparatory contours. (Preparatory contours will be discussed in section 3.5). In the intonational transcripts, I put graphic symbols in front of the orthographic representations of the syllables on which character contours begin in speech. These symbols will show which contour begins on that syllable and also indicate that the syllable so marked is majorstressed. Since these symbols indicate contour and major stress simultaneously, they are like the tonetic stress marks of many British intonation studies and can be called so. Some of them, viz. the symbols for the full fall [1], half fall [[] and rise [2] were introduced in section 1.2.2, together with the intonation phrase boundary symbol: [|]. Further symbols will be introduced below. Minor-stressed syllables will be indicated, where relevant, with []]. Unstressed syllables will not be marked. The symbols we use for the transcription enable us to show the intonation and stressing of a string of syllables within the line of written text, that is without having to draw a separate intonational diagram. Nevertheless, in this chapter, intonational diagrams (of the kind used in Chapter 1) will also be provided to help the reader visualize what the symbols represent. In later chapters, however, the intonational transcriptions will replace the diagrams in most cases, although a small number of diagrams will also be supplied for the reader’s convenience. The character contours actually appear in phonetic variants conditioned by the number of syllables over which they are spread. Since these variants are in complementary distribution, they will have no other names or symbols than the contours of which they are variants. Three such variants will be distinguished under each major-stressed contour: (a) the one-syllable, (b) the two-syllable, and (c) the threeor-more-syllable variant. These will be illustrated by the carrier phrases: Pál ‘Paul’, Mari ‘Mary’ (informal) and Angéla néni ‘Aunt Angela’. 3.2.2
Front-falling contours
The diagrams in (1), (2) and (3) illustrate the “front-falling” group of character contours.3 These include the full fall (symbol: [1]), the half fall
A Taxonomic Analysis of Hungarian Intonation
35
(symbol: [[]) and the fall–rise (symbol: [7]). The three contours can be grouped together because in their multisyllabic variants the voice radically drops down between the first and the second syllables. The full fall and the half fall are the same as Bolinger’s Profile A (Bolinger 1986: 142–9, 1989: 57). But while the full fall ends on the bottom pitch of the speaker’s normal voice range (1), the half fall ends higher, without actually going upwards (2).4 The hearer is usually able to tell whether or not the speaker has reached his/her bottom pitch (Pierrehumbert 1980: 135). This is why we regard the full fall and the half fall as two different contours. The fall–rise steps up at its end (3)c.i or rises steadily after the initial fall (3)c.ii and is similar to Bolinger’s Profile AC (Bolinger 1986: 181–3, 1989: 57).
(1) Full fall a. | `Pál. | P á
b. | `Mari. | Ma
l
c. | `Angéla néni. | An géla
ri
néni
(2) Half fall a. | [Pál. | P á
b. | [Mari. | Ma
l
c. | [Angéla néni. | An
ri
géla néni
(3) Fall–rise a. | 7Pál. | P (i)
á
l
b. | 7Mari. | Ma r
i
c. | 7Angéla néni. | An géla né An
(ii)
ni
ni géla né
The front-falling contours all indicate that the stretch of speech carrying such a contour forms a separate unit which is relatively complete and important in itself: it is neither a forward-pointer, nor a yes–no interrogative. The common meaning of all front-falling contours can be labelled as “self-contained”. It seems that the half fall has this meaning
36
Intonation and Stress: Evidence from Hungarian
alone, but the other two contours have additional components in their general meanings; viz. the full fall will also convey the meaning “finished”, and the fall–rise the meaning “conflicting”. The full fall will indicate that the stretch of speech carrying it is not only “self-contained” but also “finished”, that is it has come to an end. The fall–rise suggests that in addition to being “self-contained”, the stretch of speech carrying it is in partial contradiction with the actual or potential context. 5 The half fall, which has no other meaning than “self-contained”, implies neither finishing nor conflict: the stretch of speech carrying it is a relatively autonomous portion of the message which can be continued and suggests no conflict. For instance, if my wife asks me if it is Aunt Angela who will be looking after the children while we are away, I may use (4)a in the sense “Yes, it is Aunt Angela.” In this case the full fall will mean it is a straightforward fact and that’s all there is to say about it. On the other hand, if I choose (4)b, I imply that a further remark or explanation can be added (even if it is not actually added), for example a remark translatable as “She is very good at looking after children.” or “She always does it for us, doesn’t she?” or “At least I hope so.”, etc. (Of course, such continuations are also possible after (4)a, but there they would appear as afterthoughts.) And if I choose (4)c, the meaning of the contour will be different again: it implies that there is some kind of conflict here, with possible continuations translatable as: “But only until 8 o’clock.” or “But Uncle Joe will come too.” or “Although I haven’t asked her yet.”
(4) a. | `Angéla néni. | b. | [Angéla néni. | 6 c. | 7Angéla néni. | The front-falling contours regularly occur in statements, imperatives, question-word questions and ordinary exclamations in Hungarian. They have a great degree of formal and functional similarity with their English counterparts (apart from the difference that the fall–rise in Hungarian is not normally used as the melody of “given” information whereas it can be so used in English, see section 2.3.2). 3.2.3
Sustained contours
The next group comprises the “sustained” character contours. These include the rise (symbol: [2]), the high monotone (symbol: [4]) and the descent (symbol: [5]), when the initial syllable of the contour carrier has
A Taxonomic Analysis of Hungarian Intonation
37
some extra intensity. (When the initial syllable has no extra intensity, the sustained contours are not character contours but preparatory contours; see section 3.5.) The sustained characters correspond to various modes of Bolinger’s Profile B (Bolinger 1986: 152–5, 1989: 57–8). Their phonetic contents are shown in examples (5), (6) and (7).
(5) Rise a. | ´Pál? | (i)
l á P
b. | ´Mari? |
c. | ´Angéla néni? | ni
ri Ma
(ii)
An
é a n gél
n Angéla né
i
géla néni
(iii)
An
(6) High monotone a. | 4Pál? | Pál
b. | 4Mari? | Mari
c. | 4Angéla néni? | Angéla néni
(7) Descent a. | 5Pál? | Pá l
b. | 5Mari? | Mari
c. | 5Angéla néni? | Angéla né ni
The rise can be gradual (5)c.i, final (i.e. “end-rising”) (5)c.ii, and initial (i.e. “front-rising”) (5)c.iii. The form of the high monotone needs no comment. The descent is a narrow-ranged, gradually sloping pitch movement which starts fairly high and lacks the big drop which marks the beginning of the front-falling contours. The sustained character contours all share the meaning “forwardpointing”. These melodies show that the stretch of speech carrying them is not a yes–no interrogative, neither is it a self-contained unit
38
Intonation and Stress: Evidence from Hungarian
which is relatively complete in itself. It is rather an explicitly incomplete preparation for something complete and significant that follows or can follow in the speaker’s speech. Each contour in this group is like an index-finger which points forward to what follows or can follow. In addition to “forward-pointing”, which is the only meaning of the high monotone, the rise also has the meaning “tense”, and the descent “routine”. The gloss “tense” indicates that the stretch of speech is marked with increased tension and excitement. For instance, if I go to see Aunt Angela in her home but quite unexpectedly it is her doctor who lets me in, I may ask (8)a. “Routine” is the opposite of “tense”: the speaker considers the content of the carrier phrase a normal, natural, predictable, unexciting matter. If the doctor’s presence at the time of my visit is a matter of normal routine, I would rather say (8)c. And if I don’t want to commit myself on either interpretation, I shall probably say (8)b.
(8) a. | ´Angéla néni? | b. | 4Angéla néni? | c. | 5Angéla néni? |
The sentences in (8) are “complementary questions” (in the sense of Bolinger 1957–8) and can be regarded as topics of sentences whose comments have been ellipted. (For the notions of topic and comment, see section 1.2.3) Possible full sentence versions include Angéla néni hol van? ‘And where is Aunt Angela?’ (literally: *‘Aunt Angela where is?’), from which the comment hol van has been ellipted and only the topic Angéla néni has been retained. Sustained character contours typically accompany complementary questions and certain non-final sentence constituents in Hungarian. They are remarkably similar to their English counterparts in both form and function. 3.2.4
End-falling contours
The third group contains the “end-falling” character contours. These are the rise–fall (symbol: [6]), the monotone-fall (symbol: [¬]) and the descent-fall (symbol: [)]). These correspond to Bolinger’s Profile CA (Bolinger 1989: 58), but the A part (i.e. the falling part) in them invariably starts on the penultimate syllable when their carrier phrase is more than two syllables long. The phonetic shapes of the end-falling contours are shown in examples (9), (10) and (11). 7
A Taxonomic Analysis of Hungarian Intonation
(9)
Rise–fall a. | 6Pál? |
(i)
39
ál P
b. | 6Mari? | ri Ma
c. | 6Angéla néni? | la
né
Angé
ni né
(ii)
Angéla
ni
géla né
(iii)
An
ni
(10) Monotone-fall a. | (Pál | Pá l
b. | (Mari? | Mar i
c. | (Angéla néni? | Angéla né ni
(11) Descent-fall a. | )Pál? | Pá l
b. | )Mari? | Ma ri
c. | )Angéla néni? | Angéla
né
ni
The actual forms that the rise–fall can take may need some explanation. This contour goes up and down in the syllable when realized in a monosyllabic utterance (9)a, but the falling part may be physically missing especially if the syllable is short or ends in a voiceless consonant. When the rise–fall has two syllables to spread over, as in (9)b, the pitch steps up between the syllables and then slides back down in the second syllable. When there are more than two syllables at its disposal, the melody rises until it reaches the penultimate syllable and then drops abruptly between the penultimate and the last syllables (9)c. The rise within the rising part of the rise–fall contour can be gradual (9)c.i, final (i.e. “end-rising”) (10)c.ii, and initial (i.e. “front-rising”) (9)c.iii. The forms of the other two end-falling contours (viz. the monotone-fall and the descent-fall) should be sufficiently clear from examples (10) and (11).
40
Intonation and Stress: Evidence from Hungarian
The end-falling contours all show that the stretch of speech carrying them is grammatically a yes–no interrogative, or at least a metaphorical extension of a yes–no interrogative (see the explanation to example (15) below). But not all yes–no interrogatives are actually used to ask for information. When they are so used, the rise–fall will be chosen because this melody seems to have the meaning “yes–no interrogative + questioning”. The descent-fall shows that, although the utterance is grammatically a yes–no interrogative, it is not used to ask for information but rather to express the speaker’s disbelief or surprise over a piece of information (or experience) which (s)he already knows but finds unexpected. The meaning of the descent-fall can be glossed as: “yes–no interrogative + exclaiming”. The monotone-fall preserves the meaning “yes–no interrogative” but is neither genuinely questioning nor clearly exclaiming – it is uncommitted with respect to these meaning components. In order to appreciate these meanings, let us suppose I am told that someone wanted to see me while I was out. If I want to know if it was Aunt Angela, I ask a genuine question in the form of (12)a. However, if I am told that Aunt Angela has been here and I find this strange because I met her in the street half an hour ago, I may intone the question in the form of (12)b. And if I am told that Aunt Angela wanted to see me, though to the best of my knowledge she is on holiday in Italy, I shall probably express my surprise or disbelief in the form of (12)c.
(12) a. | 6 Angéla néni? | b. | (Angéla néni? | c. | )Angéla néni? |
The end-falling contours are used not only in yes–no questions but also in echo question–word questions (13) and repetitive question–word questions (14). The words are kivel ‘with whom’ (literally: ‘who-with’), találkoztak ‘met-3pl’, and Angélával ‘with Angela’ (literally: ‘Angelawith’).
(13) [Context:
| `Kivel találkoztak? |] ‘Who did they `meet?’ | 6Kivel találkoztak? | ‘Who did they ´meet?’ (Echo question–word question)
A Taxonomic Analysis of Hungarian Intonation
41
(14) [Context: | `Angélával találkoztak. |] ‘They met `Angela.’ | 6Kivel találkoztak? | ‘´Who did they meet?’ (Repetitive question–word question)
It has been suggested that echo questions like (13) take end-falling contours because in underlying structure they are embedded in the yes–no question: “Did you ask . . . ?” Similarly, repetitive questions like (14) can also be looked upon as being embedded in a yes–no question: “Would you repeat . . . ?” (Hetzron 1980: 397). However, such views are problematic from the point of view of the T-model we have adopted (see section 1.2.2) because they imply that syntactic deleting transformations are possible after the syntactic component has been left. One could argue that such deletions are a phenomenon restricted enough to be allowed to happen in the course of the syntax– phonology mapping, but one could also reject this possibility and look for some other reason for the rising–falling intonation. The reason could be a pragmatic one: the speaker knows that the rising–falling intonation signals a yes–no question and (s)he simply metaphorically extends this to cases like (13) and (14), because the intended meaning is in fact that of a yes–no question: “Did you ask . . . ?”, “Would you repeat . . . ?”. The end-falling contours, which primarily signal yes–no interrogatives in Hungarian, occasionally occur in certain imperatives, too, for example:
(15) Gyere come-IMP
csak
ide
egy
percre!
just
here
a
minute-for
‘Come here for a minute.’ | 6Gyere csak ide egy percre! |
I call these “belittling imperatives” because they play down the importance of the activity asked for: they express the idea that the speaker considers the effect or result of the activity (s)he is asking for to be far more important than the activity itself. This use of the end-falling
42
Intonation and Stress: Evidence from Hungarian
contours may have evolved by analogy with their use in yes–no question requests (Fónagy 1966), as in (16):
(16) Idejönnél here-come-would-2sg
egy
percre?
a
minute-for
‘Would you come here for a minute?’
8
| 6Idejönnél egy percre? | End-falling contours are characteristically Hungarian.9 They do not occur in English, and their transfer to English yes–no questions and belittling imperatives is a very common mistake among Hungarian learners of English. 10 3.2.5 Summary of the first three groups of character contours There is a significant parallel between the character contours of the first three groups, presented in sections 3.2.2–3.2.4, both with respect to their shapes and their meanings. This double parallel is shown in Figure 3.1. An inspection of Figure 3.1 reveals that, as far as their shapes are concerned, each group contains contours with a (a) rising, (b) level (monotone), and (c) descending melodic part. But the parallel is present in the meanings of the contours, too. Contours with a rising stretch have the meaning components “conflicting” (i.a), “tense” (ii.a), “questioning” (iii.a). These are all the manifestations of increased openness (i.e. tension, excitement, lack of solution), which, at a more abstract level, can be summed up as [ + open]. At the same time the additional meaning components of the contours with a descending stretch (i.e: “finished”, ii.c: “routine”, iii.c: “exclaiming”) all point towards the opposite extreme, viz. increased closedness (i.e. relaxation, solution), and can be labelled as [ – open]. And the contours containing level stretches (i.b, ii.b, iii.b) do not carry well definable additional components in their meanings, they are roughly half way between their more marked counterparts; their general characteristic is that they are neither open nor closed: they are [0 open]. These findings will be important in the autosegmental treatment of contour derivation in Chapter 4, section 4.5.1. 3.2.6
The second-type descent
There are two more character contours in Hungarian that we have to account for. One of them will be called the second-type descent. This is
A Taxonomic Analysis of Hungarian Intonation
Front-falling characters
Sustained characters
End-falling characters
i.a
ii.a
iii.a
self-contained + conflicting
forward-pointing + tense
yes–no interrogative + questioning
i.b
ii.b
iii.b
self-contained
forward-pointing
yes–no interrogative
i.c
ii.c
iii.c
self-contained + finished
forward-pointing + routine
yes–no interrogative + exclaiming
Figure 3.1
43
The first three groups of character contours
phonetically identical with the descent shown in section 3.2.3 and will have the same symbol as that [5].
(17) Second-type descent a. | 5Pál! | P ál
b. | 5Mari! | Mari
c. | 5Angéla néni! | Angéla néni
However, the meanings of the two descents differ so radically that we cannot think of them as being the meanings of one and the same character contour. It is better to analyse them as the meanings of two separate contours which happen to coincide in their surface form, just as the words writer and rider may coincide in certain American English dialects. This means that rather than recognizing one polysemous descent we shall recognize two homophonous ones.
44
Intonation and Stress: Evidence from Hungarian
The meaning of the second-type descent can be glossed as “selfcontained + evaluatively exclamatory”. The stretch of speech carrying this contour becomes an exclamation showing that the speaker is positively or negatively impressed. For example, if I am carried away by Aunt Angela’s looks or deeds, I may exclaim as in (18)a = (17)c. While the ordinary descent could be replaced by the high monotone or even a rise without losing its meaning component “forward-pointing” (see section 3.2.3), a second-type descent cannot be replaced by a high monotone or rise. The examples (18)b, (18)c cannot possibly be evaluative exclamations:
(18) a. | 5Angéla néni! | b. * | 4Angéla néni! | c. * | ´Angéla néni! | On the other hand, the second-type descent can be replaced by any of the front-falling contours and although its “evaluatively exclamatory” nature will be lost, the other component of its meaning, viz. “selfcontained”, will be preserved. The second-type descent does not exist in English and its use for exclamations in the English speech of Hungarians is another telling sign of the speaker being Hungarian. (In addition to the transfer of the characteristic Hungarian rise–fall to English yes–no questions, which was mentioned in section 3.2.4.) 3.2.7
The stylized fall
The last character contour I distinguish in Hungarian is the stylized fall (symbol: [4-]). Since it is not yet generally recognized in the Hungarian intonational literature, its treatment here will be somewhat more detailed than that of the other contours. The description is based on Varga (1995), where this contour is first described for Hungarian. The Hungarian stylized fall shows remarkable similarity to its English counterpart (Ladd 1980: 173–9; 1996: 138; Bolinger 1989: 58; Gussenhoven 1985: 123–4). It consists of a high terrace followed by a lower terrace. In its monosyllablic variant both terraces appear on a single syllable (19)a. In the disyllabic and multisyllabic variants the lower terrace is carried by the last syllable (19)b, (19)c. Minor stresses do not influence the shape of the contour. In (19)c below, the syllable nécan be minor-stressed, but the lower terrace is still restricted to the final syllable. It is here that the Hungarian stylized fall differs slightly from its English counterpart. If the English stylized fall contains a
A Taxonomic Analysis of Hungarian Intonation
45
minor-stressed syllable, the second terrace will be realized not on the last syllable alone but on the sequence of syllables starting with the minor-stressed one (cf. Gussenhoven 1985: 123–4; Hammond 1999: 157). If the stylized fall is the last (i.e. rightmost) intonation contour of an utterance, the last syllable of the carrier phrase may lengthen considerably. (And so can the last-but-one syllable, too, when it contains a phonologically long vowel, (19)c. In the examples lengthening is indicated by doubling the vowel letter.
(19) Stylized fall a. | 4-Pál! | Pá
b. | 4-Mari! |
ál
Ma rii
c. | 4-Angéla néni! | Angéla néé
11
nii
The examples in (19) have been vocatives. For instance, example (19)c is typically a routine call used by a speaker who is not in eye-contact with Aunt Angela. The contour is the same as the “vocative chant” of English descriptions (Leben 1976; Liberman 1978). Vocatives belong to a group of utterances that have a non-imperative and non-interrogative form and an imperative force. We can regard such utterances as “declarative imperatives”. Declarative imperatives include other kinds of utterances, which can also take the stylized fall when there is no eye contact between the speaker and the addressee. For instance, greetings are also declarative imperatives. If you enter your neighbour’s home but you can’t see anybody there, you may call:
(20)
Jó estét! good evening-ACC ‘Good evening.’ | 4-Jó estét! | Jó es
téét!
Declarative imperatives also include various kinds of announcements. For instance, you can use (21)a and (21)b to wake someone up, (21)c and (21)d to warn someone that the light has been left on and should be switched off, or that the phone is ringing and should be answered:
46
Intonation and Stress: Evidence from Hungarian
(21) a. Ébreszto! waken-er ‘Wake up.’ | 4-Ébreszto! | Ébresz too!
b.
d. c. Lámpa! lamp ‘Switch off the light.’ | 4-Lámpa! | Láám paa!
Nyolc óra! eight o'clock ‘It’s 8 o’clock.’ | 4-Nyolc óra! | Nyolc ó raa! Telefon! telephone ‘The phone’s ringing.’ | 4-Telefon! | Tele foon!
The following examples are informal ways of calling people to dinner:
(22) a. Vacsora! dinner ‘Dinner.’ | 4-Vacsora! | Vacso raa!
b. Kész a vacsora! ready the dinner ‘Dinner’s ready.’ 12 | 4-Kész a | 4-vacsora! | Kész vacso a raa!
But the Hungarian stylized fall can appear not only in declarative imperatives but also in the boasting, teasing, challenging calls of small children:
(23) a. Elso vagyok! first am ‘I’m the first.’ | 4-Elso vagyok! | Elso va gyook!
b. Kiskutyát kaptam! puppy-ACC received-1sg ‘I was given a puppy!’ | 4-Kiskutyát kaptam! | Kiskutyát kap taam!
c. Úgyse tudsz megfogni by-no-means can-2sg PERF-catch-INF ‘You can’t catch me.’ | 4-Úgyse tudsz megfogni! | Úgyse tudsz megfog nii!
A Taxonomic Analysis of Hungarian Intonation
47
Such infantile boasts constitute the second area of Hungarian where the stylized fall is used. The meaning of the Hungarian stylized fall can be glossed as “selfcontained + routine + mobilizing”. This suggests that the message is a relatively complete, unexciting, everyday sort of thing, but at the same time it urges the hearer to do something. The routine component is especially well demonstrated in (24). The stylized fall is appropriate in (24)a because it is used for announcing an everyday domestic event.
(24) a. Gyerekek! Kezdodik a tévé! children starts the TV ‘The TV programme is starting, children.’ | 4-Gyerekek! | 4-Kezdodik a | 4-tévé! | Gyere Kezdodik téé keek! a véé! But the same contour would hardly be possible in (24)b, where felrobbant means ‘exploded-3sg’.
(24) b. Gyerekek! Felrobbant a tévé! ‘The TV-set has exploded, children.’ * | 4-Gyerekek! | 4-Felrobbant a | 4-tévé! | Gyere Felrobbant téé keek! a véé!
The contour contains the routine component even in childish boasts. When children choose a stylized fall for their boast, they pretend that the extraordinary thing they announce (e.g. having received a new dress) is merely routine for them. The stylized fall can be replaced by any of the front-falling contours, in which case it will lose all its meaning components apart from “selfcontained”. 3.2.8
Minor stresses do not affect the character contours
At this point it is worth reminding ourselves of what was said about minor stresses in section 3.2.1, viz. that they do not initiate new character contours. Consequently, if a minor stress []] occurs in a character contour, it will not affect the shape of that contour and will not start a new contour. For instance, in all our previous examples containing the phrase Angéla néni ‘Aunt Angela’, in which there was a major stress on
48
Intonation and Stress: Evidence from Hungarian
the syllable An-, there was a possibility of the syllable né- being minorstressed, although we did not show this in the transcriptions. In any case, melodically these phrases with minor stress would not have differed from the same phrases without minor stress. In the following example there is a minor stress on the syllable é-, but this does not interrupt the rising–falling contour which started on the major-stressed syllable ta- before it.
(25) Talákoztál az édesanyámmal? met-2sg the sweet-mother-my-with ‘Did you meet my mother?’ | 6Találkoztál az ]édesanyámmal? | nyám
Talál
3.3
desa az é koztál
mal?
The appended contour
The only meaningful intonation contour left now is the appended contour. This is a low descent or monotone (being or ending at the bottom pitch of the speaker), and it contains no major stress. It is comparable to a particular mode of what Bing (1980: 23) calls the Class 0 Contour (without the optional final rise of the latter). Unlike the others, however, the appended contour is not a character contour, because (a) it differs in shape from all the character contours in that it completely lacks a high pitch, (b) it need not start with a syllable that has extra intensity, and (c) it cannot occur on independent utterances. So, however tempting it would be to regard the appended contour as another character contour, we cannot do so. The appended contour is not prominence lending in the sense that character contours are. If the first syllable of the phrase carrying an appended contour has extra intensity, that syllable is merely minor-stressed. The symbol of the appended contour [3] does not indicate any stress on the syllable before which it stands, it only indicates the start of the contour. In (26) the appended contour accompanies the italicized final vocatives. The new words are: maga ‘you’ (formal, like French vous) and az ‘that’ (demonstrative). The symbol [-] represents an audible pause (of any length) and []] represents minor stress. In the examples of (26) both the pause and the minor stress are optional.
A Taxonomic Analysis of Hungarian Intonation
49
(26) Appended contour a. | 6Maga az, | - 3 ]Pál? | ga Ma az, Pál?
b. | 6Maga az, | - 3 ]Mari? | ga Ma az, Mari?
‘Is that you, Paul?’
‘Is that you, Mary?’
c. | 6Maga az, | - 3 ]Angéla néni? | ga Ma az, Angéla néni? ‘Is that you, Aunt Angela?’ In these examples the appended contour constitutes a separate intonational phrase, whether or not there is a pause before it, because it is melodically separated from the preceding rising–falling contour. After a half fall or full fall, however, the appended contour is not melodically separated and so it can only retain its identity if it is separated from the end of the preceding fall by an audible pause. So while the quoting clause mondta takes an appended contour and forms a separate intonation phrase in (27)a, the same word mondta is part of the full fall in (27)b. The words in the examples are: a ‘the’; nagymama ‘grandma’; mondta ‘said-3sg’.
(27) a. | A 1nagymama, | - 3 mondta. | nagy A mama, mondta. ‘Grandma, – he said.’ b. | A 1nagymama mondta. | nagy A mama mon dta. ‘Grandma said it.’
50
Intonation and Stress: Evidence from Hungarian
The minimal pair in (27) reveals why the pause and the lack of pause are prosodemes: the presence or absence of pause may be the only physical difference between the two utterances. In (26)a, b and c the appended contour was carried by a vocative, in (27)a by a quoting clause. Vocatives and quoting clauses belong to “inorganic blocks”, which will be explored in section 5.3.2. When a pause interrupts a fully falling character contour, which continues unchanged after the pause, and the post-pausal part of the continued contour cannot be equated with an appended contour because syntactically its carrier phrase is not an inorganic block, the pause is not regarded as an intonational phrase boundary. For instance, in (28) the whole utterance carries a full fall, which is interrupted in the middle by an obligatory pause, but the post-pausal part is not an appended contour, and so the utterance is one single intonational phrase.
(28) Nem a barátomról, Sándorról beszéltek. not the friend-my-about Alexander-about spoke-3pl ‘It’s not my friend, Sándor, that they were talking about.’ | 1Nem a barátomról, - ]Sándorról beszéltek. | Nem a barátomról, Sándorról beszéltek.
The situation was different in (27)a, because the pause there was not merely an interruption but in fact initiated an appended contour, that is a new intonation phrase. The meaning of the appended contour can be glossed as: “selfcontained + finished + inorganic”. The last label means that the stretch of speech carrying this melody constitutes an inorganic part of the utterance and has a low information value: its function is maintaining contact between speaker and hearer and is used chiefly in utterancefinal “class 0 expressions” (Bing 1980: 21–52), that is vocatives, quoting clauses, etc. at the end of the utterance.
3.4
The intonational lexicon
We can now summarize the intonational lexicon of Hungarian as shown in Table 3.2: 13
A Taxonomic Analysis of Hungarian Intonation Table 3.2
51
The intonational lexicon of Hungarian
1. Full fall 2. Half fall 3. Fall–rise
self-contained + finished self-contained self-contained + conflicting
4. Rise 5. High monotone 6. Descent
forward-pointing + tense forward-pointing forward-pointing + routine
7. Rise–fall 8. Monotone-fall 9. Descent-fall
yes–no interrog. + questioning yes–no interrog. yes–no interrog. + exclaiming
10. 2nd-type descent 11. Stylized fall 12. Appended contour
self-contained + evaluatively exclamatory self-contained + routine + mobilizing self-contained + finished + inorganic
The intonational lexicon in Table 3.2 is at the same time a taxonomic list of meaningful Hungarian intonation contours, established on the basis of surface contrasts. The first eleven entries in the lexicon are character contours, two of which (Nos. 6 and 10) coincide in phonetic shape (i.e. are intonational “homonyms”). The number of meaningful contours may appear to be rather high but, in reality, it is not high if we compare it with the tens of thousands of words in verbal lexicons, or even with the 30–40 vocal signals present in some animal communication systems. These contours occur with different frequencies. The two recordings in my corpus, a 50-minute long narrative read aloud and a 20-minute long spoken debate (section 1.3), yield the percentages shown in Table 3.3. In order to produce these percentages, the dialogues within the narrative and the whole of the spoken debate were brought together as “spontaneous speech” (1,924 meaningful contours). The non-dialogue parts of the narrative have been labelled “reading aloud” (1,824 meaningful contours). The descending and second-type descending contours have been collapsed. Although a larger and more varied corpus may provide more exact data, some tendencies are quite conspicuous already: (i) The half fall clearly emerges as the contour with the highest count in both categories but it has a higher proportion in spontanous speech than in reading aloud.
52
Intonation and Stress: Evidence from Hungarian
Table 3.3 The frequency of occurrence of meaningful contours in the corpus examined Spontaneous speech
Full fall Half fall Fall–rise Rise High monotone Descent Rise–fall Monotone-fall Descent-fall Stylized fall Appended Total
Reding aloud
Total
No.
%
No.
%
No.
%
452 1,152 32 134 55 66 30 3 0 0 0 1,924
23.49 59.88 1.66 6.96 2.86 3.43 1.56 0.16 0.00 0.00 0.00 100.00
416 886 14 227 94 156 1 0 0 0 30 1,824
22.81 48.57 0.77 12.44 5.15 8.55 0.05 0.00 0.00 0.00 1.64 99.98
868 2,038 46 361 149 222 31 3 0 0 30 3,748
23.16 54.37 1.23 9.63 3.97 5.92 0.83 0.08 0.00 0.00 0.80 99.99
(ii) There is a significantly higher number of sustained contours (rises, high monotones, descents) in reading aloud than in spontanous speech. (iii) The rise–fall is a characteristic feature of spontaneous speech rather than of reading aloud. (iv) The appended contour is a characteristic feature of reading aloud rather than of spontaneous speech.
3.5
The preparatory contour
The twelve contours we have examined so far constitute the so-called intonational lexicon of Standard Hungarian. In addition to these contours mention must also be made of a thirteenth kind of contour which, however, has no meaning of its own and is therefore no part of the intonational lexicon. This is the preparatory contour, which is the automatic (default) melody of a stretch of speech at the beginning of an intonation phrase, carrying neither a character contour nor an appended contour. It is a melody without major stress, the prosodic accompaniment of articles, conjunctions, minor-stressed or unstressed sentence constituents at the beginning of an intonation phrase. The preparatory contour always initiates a new intonation phrase. For instance, the italicized string of words és ha ‘and if’ in example (29) appears with a preparatory contour.
A Taxonomic Analysis of Hungarian Intonation
53
(29) Preparatory Contour | És ha 2Angéla néni? | néni géla An És ha ‘And [what] if it is Aunt Angela?’ The preparatory contour is either (a) a lowish monotone (but higher than the bottom pitch of the speaker), in which case it may contain syllables with extra intensity (these syllables will count as minor-stressed); or (b) it is like one of the sustained contours (rise, high monotone, descent), in which case it cannot contain a syllable that has extra intensity (and will contain only unstressed syllables). 14 In transcription, the preparatory contour needs no separate symbol because it occurs immediately after an intonation phrase boundary [|], as in (29). 15 The examples in (30) mean different things – the difference correlates with (30)a having a preparatory contour (on the word ide ‘here’) while (30)b has none. The other words are: gyere ‘come-IMP-2sg’; ne ‘don’t’; and firkálj ‘scribble-IMP-2sg’. The pause is likely but not essential.
(30) a. | `Gyere! |
-
Ide
Gye
`ne
firkálj! |
ne re!
Ide
firkálj!
‘Come. Don't scribble here.’
b. | `Gyere ide! | Gye
-
`Ne
firkálj! |
Ne re ide!
firkálj!
‘Come here. Don’t scribble.’ The preparatory contour on the word ide in (30)a is easy to identify because it is melodically separated from the preceding contour even if there is no pause between them. However, when the preparatory contour continues the melody of the preceding contour – that is it is not melodically separate from the preceding contour – two things can happen. If there is no pause before it, the preparatory contour stops being
54
Intonation and Stress: Evidence from Hungarian
a preparatory contour and is analysed as part of the contour that precedes it, as in (31)a. If there is an audible pause before the preparatory contour, it retains its identity, as in (31)b. When a character contour is interrupted by a pause after which the contour continues, but the post-pausal part can be equated with the preparatory contour of a new intonational phrase, we shall regard the pause as an intonational phrase boundary. Although the melody after the pause in such cases is a continuation of the preceding contour, it could be a different melody: the preparatory contour could take another shape as well. This is why the words vagy csak carry a preparatory contour in (31)b but not in (31)a.
(31) Ötszázat,
vagy
five hundred-ACC or
csak
hármat?
only
three-ACC
‘500 or only 300?’ a. | 4Ötszázat, vagy csak | `hármat? | Ötszázat, vagy csak hár mat? b. | 4Ötszázat, | - vagy csak `hármat? | Ötszázat
vagy csak hár mat?
The significance of the preparatory contour lies in the fact that its presence restricts the extent of the character contour or appended contour before it.
3.6 3.6.1
The unit of Hungarian intonation The intonation phrase
The unit of Hungarian intonation is the intonation(al) phrase (IP), and its boundaries are indicated by means of the symbol [|]. The obligatory part of a Hungarian IP is the terminal part, filled by any of the meaningful intonation contours that we have seen in sections 3.2–3.3, that is by either a character contour or an appended contour. All the character contours are always IP-final, that is terminal, and thus occupy the terminal part of the IP, apart from the half fall, which is IP-final only if it ends in an audible pause. The other contours in the terminal part may or may not end in a pause. So an IP may end in an audible pause, but this is obligatory only when its final character is a half falling contour. The non-final character contours are half falling contours that do not end in a pause. These constitute the scale of the IP and can be
A Taxonomic Analysis of Hungarian Intonation
55
called the scalar contours. So the scale is an optional unit of the IP. It may be preceded by an optional preparatory part, which contains a preparatory contour, with unstressed or maximally minor-stressed syllables. Hungarian IPs therefore have the following internal structure:
(32) Hungarian IP (Preparatory Part) + (Scale) + Terminal Part
16
All these possibilities are shown in (33).
(33) És
ha
and if
levették
a
off-took-3pl
the books-ACC
könyveket
a
polcokról?
the
shelves-from
‘And [what] if they took the books off the shelves?’ | És ha [levették a [könyveket a ´polcokról? | le És ha
vették a
Prep.
l?
ró
köny veket a
polcok
Scale
Terminal
IP = Utterance There are two restrictions here. The first one is that, if the terminal part is occupied by an appended contour, then neither a preparatory part, nor a scale is possible before it; see, for example, the second IP in (34).
(34) Emlékszel?
–
remember-2sg
kérdezte
Péter.
asked-3sg
Peter
‘Do you remember? – asked Peter.’ | 6Emlékszel? | 3 ]kérdezte ]Péter. | Em
lék
szel? kérdezte Péter.
Terminal
Terminal
IP
IP 17
Utterance
56
Intonation and Stress: Evidence from Hungarian
The second restriction is that a stylized fall cannot have a scale before it (though it can have a preparatory part). When a sentence is realized with a final stylized fall, the other major-stressed syllables in the sentence also initiate the stylized fall and so the sentence breaks up into as many IPs as the number of the stylized contours in it, see (35). (See also (22)b and (24) above.)
(35) Mi meg indulunk Keszthelyre! we on-the-other-hand start-3pl Keszthely-to ‘And we are leaving for Keszthely.’ | Mi meg 4-indulunk | 4-Keszthelyre! | indulunk Keszthelyree! Mi meg
Prep.
Terminal
IP
Terminal IP
Utterance 3.6.2
Why the contour is not an alternative unit of intonation
The Hungarian IP, presented in the previous section, is able to accommodate more than one pitch accent, and so it looks similar to an English IP (tone-group). 18 In English, IPs are organized around a pitch accented syllable of outstanding prominence (the “nucleus” or “tonic”), which initiates a nuclear contour, and may or may not be preceded by pitch accented syllables of less prominence (in the “head”). However, since we do not distinguish Hungarian pitch accents according to strength, it may at first seem unreasonable to try to establish IPs in Hungarian. After all, if all accents are regarded as equal, there seems to be no justification in picking out some of them as being the nuclei of IPs and it may seem correct to analyse Hungarian intonation in terms of “units” that are co-extensive with single character contours and appended contours. 19 This could be called the “contour-unit framework”, as opposed to the “IP framework”, presented in section 3.6.1. One has to admit that the contour-unit framework is a logically impeccable alternative. All we have to do is introduce a new symbol (this could be, for example, [;]) to mark the beginning of preparatory contours. Since the appended contour, the preparatory contour and the various character contours are all indicated by tonetic stress marks now,
A Taxonomic Analysis of Hungarian Intonation
57
their boundaries are clear and no separate IP boundaries are necessary in the transcription. Let us compare transcriptions of the same utterance in the two frameworks. (36)a has been constructed according to the IP framework, (36)b according to the contour-unit framework. The IP framework is recognizable by the vertical lines it uses for the IP boundaries in the transcription.
(36) E jelenlegi magyar úthálózat this present Hungarian road-network közlekedésbiztonság szempontjából nagy traffic-safety point-of-view-its-form big tömegében nem felel meg a követelményeknek. mass-its-in not answers PERF the requirements-to ‘The present Hungarian road-system, in large part, does not meet the requirements of traffic safety.’ a. | E [jelenlegi [magyar 2úthálózat | [közlekedésbiztonság 2szempontjából | [nagy tömegében [nem felel meg a 1követelményeknek. | b. ⊥E [jelenlegi [magyar 2úthálózat [közlekedésbiztonság 2szempontjából [nagy tömegében [nem felel meg a 1követelményeknek.
Although both analyses are possible, there are reasons why I think the IP analysis, (36)a, is preferable. First, however, let us refute the objection that no pitch accent can be identified around which a Hungarian IP can be organized. We can postulate IPs for Hungarian because certain character contours (viz. the ones we called “terminal” contours in section 3.6.1) can be picked out as having an analogous role to the English nuclear contours. However, they stand out not because of their extra prominence but due to their shape (or in the case of the half fall, due to the accompanying pause). Let us now consider the advantages of the IP approach over the contour-unit approach. A fairly obvious advantage of the IP approach is that it is capable of showing the unity of syntactico-semantic constituents as well as their separation from other syntactico-semantic constituents along the syntagmatic axis, and thus they form “sense-units”. For instance, in (36)a each of the first two IPs contains a topic constituent, and the third IP corresponds to the comment of the sentence. (For the notions of topic and comment, see section 1.2.3.) By contrast, the
58
Intonation and Stress: Evidence from Hungarian
contour-unit approach is too atomistic, and is unable to show such syntactico-semantic divisions, as is clear from (36)b. But this advantage of the IP approach, that it can signal sense-units, seems to dissolve when the IPs do not coincide with such units, as, for example, in (37)a:
(37) Megvették a szomszéd házát. PERF-bought-3pl the neighbour house-his -ACC ‘They bought their neighbour’s house.’ a. |[Megvették a 4szomszéd| 1házát.| b. [Megvették a 4szomszéd 1házát.
In such cases the contour-unit framework (37)b may seem preferable, because it does not suggest any division into syntactico-semantic constituents. However, cases like (37) can be looked upon as being derived from a more basic intonational segmentation of the utterance in which the IPs are more closely related to sense-units. They are cases of attitudinal contour replacement (“Replacement of Half Fall by High Monotone”), which we shall discuss in Chapter 5, section 5.4.7. I find that examples such as (37)a do not provide sufficient evidence against the IP approach. A second piece of evidence in support of the IP analysis is related to the phenomenon of downdrift (see the next section). Downdrift means that certain character contours become successively lower within an IP, and that the IPs become successively lower within an utterance, though each IP starts a fresh process of IP-level downdrift. Consider (38):
(38)
Megkérdeztem, hogy mennyit fizessek, PERF-asked-lsg that=CONJ how-much-ACC pay-shall-lsg és azt mondta, hogy semmit. and that=DEM-ACC said-3sg that=CONJ nothing-ACC ‘I asked him how much to pay, and he told me to pay nothing.’ |[Megkérdeztem, hogy [mennyit [fizessek, | - és [azt mondta, hogy 1semmit.| Meg kérdeztem, hogy men fi azt sem nyit és mondta, hogy zessek, mit.
A Taxonomic Analysis of Hungarian Intonation
59
In (38), both IPs display internal downdrift so that the second IP starts downdrift anew (on the syllable azt, which is at roughly the same height as the syllable fi-), but at the same time there is utterance-level (inter-IP) downdrift too; so that the second IP starts somewhat lower than the first. Such complex downdrift phenomena can be adequately treated only in an IP framework. Related to downdrift is the problem of the Strict Layer Hypothesis. Division of the Hungarian utterance into IPs is in accordance with the Strict Layer Hypothesis, advocated by Selkirk (1984: 26). According to this hypothesis, utterances are divided into IPs exhaustively and no IP can contain another IP. This has been challenged by Ladd (1986), who says that an English IP may contain another IP, after which the matrix IP continues the downdrift started before the interruption, as in (39).
(39) [IP The book on the table [IP it seems to me] was a gift from my mother.] (adapted from Ladd (1896: 324)) In Hungarian, also, it is possible to insert an utterance in the middle of a larger utterance, after which the larger utterance continues as if it had not been interrupted, as in (40):
(40) Találkoztam – jól ülsz? – az Angélával! met-lsg well sit-2sg the Angela-with ‘I’ve met – are you sitting comfortably? – Angela.’ 20 éá l val! | | [Találkoztam | - – 6jól ülsz? – | - az 1Ang
IP
IP
IP
Utterance Nevertheless, such discontinuous utterances can be analysed perfectly well in terms of IPs linearly following one another, because the relatively lower peak of the third IP can be explained by utterance-level (inter-IP) downdrift. So I can see no compelling reason for dismissing the Strict Layer Hypothesis in Hungarian. In a contour-unit framework, where the only recognized grouping of character contours is the utterance itself, there is no way to adequately account for downdrift.
60
Intonation and Stress: Evidence from Hungarian
A further argument in favour of the IP analysis will be provided in Chapter 8, where it will be shown that the domain of rhythmical secondary stress assignment is also the intonation phrase. 3.6.3
More on the intonation phrase
We have seen that the unit of Hungarian intonation can be either the IP or the contour-unit. The important difference between them is that the contour-unit framework identifies intonational units with individual meaningful contours, whereas the IP framework allows the intonational units to contain groups of meaningful contours arranged in a certain pattern. Although both approaches are suitable for notational purposes, certain phenomena can be adequately accounted for only by the IP approach and therefore it is the IP that has to be recognized as the unit of Hungarian intonation. This unit is identified on the basis of the characteristic internal structure that it has, and it is derived by cutting up the utterance rather than by putting together phonological phrases. In other words, it is defined not with reference to lower-ranking prosodic units which are treated rather heterogeneously in the literature and which are difficult to identify, but with reference to the highest-ranking prosodic unit, the utterance, whose boundaries are usually clear. We shall examine the ways in which Hungarian IPs are formed in Chapter 5. Our IP is different from Vogel’s intonational phrase (Vogel 1987: 245; cf. Nespor and Vogel 1986, and also, mutatis mutandis, Hunyadi 1999), which is built from below, that is from phonological phrases, and is defined in such a way that logical form is allowed to play a role in the definition. Vogel’s intonational phrase is said to be the domain of l-palatalization and of certain kinds of stress reduction in Hungarian (Vogel and Kenesei 1987; Kenesei and Vogel 1989). However, it is quite obviously not a unit of intonation in the normal sense of the term, because it has neither the characteristic internal structure nor the audible boundaries that intonation units can have, and which intonationists regard as crucial for the definition of IPs (cf. Ladd 1996: 235–51). Therefore I think Vogel’s IP should be regarded as a “super phonological phrase” rather than a real IP, a unit which, important though it may be in other respects, is not needed for the purposes of the present analysis. The preparatory part, scale and terminal part within the Hungarian IP are not prosodic constituents. The prosodic constituents that our IP can be divided into are the accent phrase (AP), the foot (Ft) and the syllable. The foot lasts from one syllable with extra intensity to the next, or – if there is no further syllable with extra intensity – to the end of the IP. Our IP consists of a certain number of feet plus an optional unfooted
A Taxonomic Analysis of Hungarian Intonation
61
clitic, corresponding to the preparatory part, at its left edge. Majorstressed feet, together with all the minor-stressed feet on their right, can be said to constitute “accent phrases” (Cruttenden 1997: 24). An accent phrase is the domain of a pitch accent, that is the domain of a character contour. (Consequently, it does not occur in an IP which only contains an appended contour.) For instance, (41) is an IP which consists of an unfooted clitic and three accent phrases. The unfooted clitic at the beginning constitutes the preparatory part of the IP, the first and second accent phrases form the scale, and the third accent phrase is the terminal part of the IP.
(41) De eredetiben olvasta az angol regényeket. but original-in read-3sg the English novels-ACC ‘But he read the English novels in the original.’ IP: | De [eredetiben ]olvasta az [angol 1regényeket. | AP: | . . . ( ( ( | ( Ft: | . . . ( ( ( |
3.7
Downdrift and upstep
The major stressed syllables within the scale of the IP are normally subject to downdrift (also known as declination): each of the consecutive scale contours normally starts at a slightly lower pitch than the preceding scale contour. This is an automatic, mechanical consequence of the speaker gradually running out of air as (s)he proceeds with the realization of the IP. It is illustrated by (42) below, where the starting pitch of könyveket ‘books-ACC’ is somewhat lower than that of levették ‘off-took3pl’, but is slightly higher than the starting point of polcokról ‘shelvesfrom’. Downdrift within the IP also regularly extends to the terminal part, if the terminal contour is front-falling, as in (42)a, or end-falling, as in (42)b. In (42)a and (42)b the peak of the terminal contour is lower than it would be if it stood at the beginning of an IP.
(42) a. | [Levették a [könyveket a 1polcokról. | Le vették a
köny pol veket a cokról.
‘They took the books off the shelves.’
62
Intonation and Stress: Evidence from Hungarian
b. | [Levették a [könyveket a 6polcokról? | Le vették a
köny cok veket a pol ról?
‘Did they take the books off the shelves?’ But if the terminal contour is a sustained contour, it does not normally share the feature of downdrift with the scale. For instance, the rising contour on polcokról in (33) above is not perceptibly downdrifted. At the beginning of a new IP, downdrift starts anew, but in some cases downdrift can be observed even between consecutive IPs, if they belong to the same utterance. Such utterance-level (inter-IP) downdrift can be observed between the IPs of an utterance in the following circumstances: the terminal contour of the affected IP is front-falling, as in the second IP of (43) and the third IP of (45); or end-falling, as in the second IP of (44); or identical with the terminal contour of the preceding IP, as in the second IP of (45).
(43) Kérsz kávét? Angéla? want-2sg coffee-ACC Angela ‘Would you like some coffee, Angela?’ | 6Kérsz kávét, | 1Angéla? | ká Kérsz An vét, géla? (44) Angéla, kérsz kávét? ‘Angela, would you like some coffee?’ | 1Angéla, | 6kérsz kávét? | An ká géla, kérsz vét? (45) Káposztát, karalábét, karfiolt. cabbage-ACC kohlrabi-ACC cauliflower-ACC ‘Cabbage, kohlrabi, cauliflower.’ | 2Káposztát, | 2karalábét, | 1karfiolt. | tát, Káposz
lábét,
kara
kar fiolt.
A Taxonomic Analysis of Hungarian Intonation
63
Where downdrift is possible, the deliberate suspension of downdrift is also possible; this will be called upstepping. Between contours of the front-falling type (see section 3.2.2), between contours of the end-falling type (see section 3.2.4) or between identical sustained contours (see section 3.2.3) the upstep can take two forms: either the peak of the second contour is at the same pitch height as the peak of the first (equalizing upstep), or the peak of the second contour is higher in pitch than the peak of the first (exceeding upstep), as in (46)b and (46)c, respectively. (46)a shows the downdrifted version. While downdrift has no special symbol in our intonational transcriptions, the symbol of the exceeding upstep will be [ 8 ] and the symbol of the equalizing upstep will be [ = ], both placed before the syllable to which they apply.
(46) Vedd fel a másikat! take-IMP-2sg up the other-ACC ‘Pick up the other one.’ a. | [Vedd fel a 1másikat! | Vedd má fel a sikat! =
b. | [Vedd fel a 1másikat! | Vedd má fel a si kat! 8
c. | [Vedd fel a 1másikat! | má Vedd fel a si kat! The utterances in (46)a, (46)b and (46)c are all attitudinally different. While (46)a is a simple imperative, (46)b suggests that the speaker regards the proposal as self-evident and (46)c conveys not only that the speaker regards the proposal as self-evident but also that (s)he is surprised or irritated that such an obvious proposal has not occurred to his/her partner. In other combinations, that is where the two contours are not identical or not both front-falling or not both end-falling, the upstep of the second contour can only be of the exceeding kind, as in (47). This means that the affected contour appears with the peak height that it would have at the beginning of an utterance.
64
Intonation and Stress: Evidence from Hungarian
(47) Angéla, kérsz kávét? (cf. (44)) ‘Angela, would you like some 8 | 1Angéla, | 6kérsz kávét? | ká
An géla,
3.8
coffee?’
kérsz
vét?
Summary
This has been a key chapter. We have carried out a taxonomic phonological analysis of the prosodemes of Hungarian: the character contours; the appended and preparatory contours; the stress, pause and upstep prosodemes. We have identified prosodemes on the basis of their ability to establish surface contrasts. We have established an intonational lexicon of melodic prosodemes with twelve entries, and described the internal structure of the Hungarian intonation(al) phrase (IP). We have also introduced a system of graphic symbols for the notation of stress and intonation.
4 An Autosegmental Analysis of Hungarian Intonation
4.1
Introduction
In this chapter we reanalyse in autosegmental terms the intonation contours which are the entries in the Hungarian intonational lexicon (see section 3.4), and give each of them an autosegmental representation in a framework established along the lines suggested by Gussenhoven (1983a, 1985). The justification for an autosegmental account of intonation lies in the existence of speech melodies that are independent of the segmental make-up of the carrier phrases, and in the possibility of associating these melodies with the syllables of the carrier phrases in different ways.1 An autosegmental phonological representation consists of several autonomous tiers running parallel to each other. From an intonational point of view these are the tier of syllabic representation and the tier of tonal representation. The tonal tier contains the intonation contours (melodic prosodemes) represented by various sequences of melodic elements, that is by sequences typically consisting of H (high) and L (low) tones. The melodic elements are tied to the appropriate syllables of the syllabic tier by means of association lines, according to a number of association conventions and rearrangement rules. The chapter first compares the present account with an earlier autosegmental analysis of Hungarian intonation (section 4.2). It then presents the representational conventions of the present analysis (section 4.3), and shows how the twelve meaningful intonation contours identified in Chapter 3 can be derived from the representations of three basic contours (section 4.4). The rule-governed process of deriving intonation contours from basic contours is called melody formation. This is the topic of section 4.5. Melody 65
66
Intonation and Stress: Evidence from Hungarian
formation is a process in which derived intonation contours are produced; it is a morphological kind of process, analogous to word formation (see Gussenhoven 1983a, 1985). The basic contours correspond to simple words (free root morphemes) and the derived contours to derived words (in which derivational affixes are added to root morphemes or to previously formed derivatives). Melody formation takes place in two phases. In the first, the primary derivational rules affect all the basic contours. In the second, the secondary derivational rules change particular contours. Between the two phases (i.e. before the secondary derivational rules apply) the representations of the intonation contours are simplified. The derivational rules affect meaning and are optional in the sense that they are not phonologically determined. The simplifying rules do not affect meaning and are obligatory. Two types of horizontal double arrows are used in the formulation of rules: =obl⇒ introduces obligatory changes and =opt⇒ introduces optional ones. The rules and representations are summarized in section 4.6. Elements of the contour representations are associated with the syllables of the carrier phrase. This process is governed by association conventions, proposed in section 4.7. The associated representations are then submitted to certain rearrangement rules (section 4.8). Features of the phonetic realization are presented in section 4.9. Finally the Hungarian intonation phrase is revisited, this time in an autosegmental framework, in section 4.10. This chapter is a revised version of the relevant part of Varga (1994).
4.2
The Kálmán–Kornai analysis
The pioneering work in the autosegmental analysis of Hungarian intonation was done by Kálmán and Kornai (1985) and Kornai and Kálmán (1989), and we shall refer to it as the Kálmán–Kornai analysis. This analysis starts by positing a basic melody which is represented as LHL, from which certain pragmatically-guided rules derive a fall (HL), a fall–rise (HLH), a high monotone (H) and a rise–fall (LHL). The first difference between the Kálmán–Kornai analysis and ours is that the former did not deal with the formal and semantic relations that exist between contours. In contrast, our analysis concentrates on these relations, regarding the contours as entries in an intonational lexicon, that is as elements that are analogous to words, having their own derivational “morphology”. Another difference between the two approaches is that we are going to recognize as phonologically distinct certain contours which the
An Autosegmental Analysis of Hungarian Intonation
67
Kálmán–Kornai analysis treated as phonetic variations. According to Kálmán and Kornai (1985) the half fall and the full fall are the same contour and have the same representation (HL), the difference between them being due to a low-ranking phonetic rule, viz. tonal assimilation. Tonal assimilation ensures that the Ls between two Hs become higher within the same utterance. That is to say, the half fall is a positional variant of the full fall: the full fall is realized as half fall (with a higher L) when it is followed by a melody starting with H in the same utterance. This, however, does not explain half falls that stand in utterance-final position, where no melody starting with H follows them, as in (1):
(1) Furcsa ez a strange this the ‘This milk is strange.’ | [Furcsa ez a tej... |
tej... milk
One could object that (1) does have a continuation in its underlying syntactic representation and the melody of this continuation begins with H. (It may be, for example, another HL.) The continuation turns the melody before it into a half fall and is then deleted. This objection presupposes that some syntactic deletions are carried out after melodic assignment.2 Such views were not unheard of in earlier periods of generative linguistics (e.g. Pope 1971), but the real problem is that explanations based on them may turn out to be wrong. Take, for example, the “offering” kind of complementary question:
(2)a. | 2Konyakot? | ‘Brandy?’ brandy-ACC Syntactically, such questions can be considered as the remaining topics of yes–no questions whose comment has been deleted. (2)a probably derives from (2)b, in which konyakot is the topic and tölthetek is the comment:
(2)b. | [Konyakot 6 tölthetek? | ‘May I pour you some brandy?’ pour-can-lsg Intonationally, however, a similar explanation is unfeasible. The rising contour of (2)a cannot be the melody of a remaining topic because the
68
Intonation and Stress: Evidence from Hungarian
topic of a yes–no question never has a rising contour; it typically has a half fall, as in (2)b. Consequently the melody of (2)a cannot possibly derive from the melody of the topic in (2)b. What is more likely is that (2)a is assigned a rising melody after the syntactic operation has produced (2)a from (2)b. A different kind of problem arises in connection with the descending contour. For Kálmán and Kornai (1985) this is only a phonetic variant of the high monotone when the latter is pronounced with downdrift. It is true that in some cases the high monotone can be replaced by a descent, but this change in form causes a change in attitude (see section 3.2.3), and so it can be the result of a conscious choice rather than a phonetic accident. While the intonation of (3)a is appropriate to the dramatic content of the utterance, (3)b sounds routine and unduly light-hearted, as if to suggest something like “who cares?”.
(3) És ha amputálni kell? and if amputate-INF necessary ‘And [ what ] if it has to be amputated?’ a. | És ha 4amputálni kell? | b. | És ha 5amputálni kell? |
In addition, the contour we called “second-type descent” (see section 3.2.6) can never be replaced by the high monotone and so it can never be derived from the latter; see (4)a and (4)b.
(4) Hogy megnott ez a how PERF-grew-3sg this the ‘How this girl has grown!’ a. * | 4Hogy megnott ez a kislány! | b. | 5Hogy megnott ez a kislány! |
kislány! small-girl
So the descent (and especially the second-type descent) cannot be regarded as a phonetic variant of the high monotone. In sum, the Kálmán–Kornai analysis, although attractive and elegant, oversimplifies the system of Hungarian character contours and does not deal with the genetic relations between them. Therefore, an alternative analysis is justified. The rest of the chapter is devoted to one such analysis.
An Autosegmental Analysis of Hungarian Intonation
4.3
69
The representational conventions
In the melodic representations I show the end of a melody by the contour end symbol: $. This is not to be associated with any syllable. Among the melodic elements we recognize not only the H and L tones but also the “lowered high” (H9), “lowered low” (L9), “raised high” (H8) and “raised low” (L8) tones. These lowered and raised tones occur in derived melodies and can only be interpreted if the melodic representation in which they occur contains the corresponding non-lowered and non-raised tone as well. The raised high tone (H8) is slightly higher than the high tone (H) within the same melody, the lowered high tone (H9) is slightly lower but not low. The raised low tone (L8) is slightly higher than the low tone (L) occurring in the same melody: it phonetically coincides with the lowered high tone (L8 = H9). The lowered low tone (L9) is somewhat lower than the low tone (L) of the same melody: it corresponds to the bottom pitch, that is the lower limit of the normal voice range of the speaker.3
4
(5) Tones and the corresponding pitch levels (Preliminary version) H8 H H9 = L8 L L9 The raised and lowered tones come into being by interference from the tones on an ancillary tonal tier (see Inkelas and Leben 1987). The ancillary tones are associated with tones, not with syllables, and they do not spread. The Greek letter σ (sigma) is employed as the symbol for the syllable (Table 4.1). In addition to the elements presented so far, we shall also recognize a so-called “stylizer” (symbolized as + ), occurring in the representation of Table 4.1
Tones and their representations
Syllabic tier Tonal tier Ancillary tonal tier
L9
L
L8 = H9
H
H8
σ | L | L
σ | L
σ σ | | L H | | H L
σ | H
σ | H | H
70
Intonation and Stress: Evidence from Hungarian
the stylized fall (see section 4.5.2). The stylizer, like the contour end, is not a melodic element and is not to be associated with syllables. In the melodic representations the initial tone of a character contour (which is to be aligned with the major-stressed, initial syllable of the string of syllables that carries the contour) will be followed by a star, for example H*, and the melodic elements belonging to the same contour will be separated from each other by a dot, for example the full fall will be represented as: H*.L.L9$.
4.4
The basic contours
Analysing the forms and meanings of the twelve entries in the Hungarian intonational lexicon, I presume that they can be derived from three basic contours. These are the following:
(6) a. H*.L.L$ b. H*.H$ c. H*.H.L$
(half falling contour; [x) (high monotone contour; 4x) (monotone-falling contour; (x)
These three melodies have been chosen to be the basic contours because it is these melodies for which the derivational rules can be formulated in the most economical way. They can all be seen as expansions of the starred tone H*. They contain a maximum of three tones, of which two adjacent tones are identical. They do not obey the Obligatory Contour Principle (OCP, Leben 1973). The OCP would reduce a series of identical tones to a single tone, so that (6)a and (6)c would equally become H*.L$, while (6)b would become H*$. The OCP is not observed here because we want to avoid the coincidence of (6)a and (6)c, and also because the primary derivational rules are easier to formulate if they can apply to unsimplified representations.5 Later, however, when the primary derivational rules have already been applied, the Identical Tone Reduction Rule (a limited version of OCP) will simplify all the basic contours except (6)c. This rule is given in (7). (T stands for any tone, Tα stands for any tone, provided it is the same tone within the same rule.)
(7) Identical Tone Reduction Rule (limited version of OCP) .Tα =obl: 0 / Tα $ This rule collapses the identical tones into one single tone before the contour end ($) in such a way that it deletes the second of the two
An Autosegmental Analysis of Hungarian Intonation
71
identical tones. Consequently, the basic contours will be simplified in the following way:
(8) a. H*.L.L$ =obl: H*.L$ b. H*.H$ =obl: H*$ The third basic contour (H*.H.L$) will not be simplified, because the identical elements in it are not immediately before the contour end and so they do not trigger (7). If we also simplified the representation H*.H.L$ into H*.L$, we would have no way of showing the difference between (9)a and (9)b, although (9)a and (9)b are different contours.
(9) a.
b. H
H
L
L te.
És
És
H*.L$ ‘And you.’
te?
H*.H.L$ ‘And you?’
This is why we had to formulate (7) in such a way that it should not be able to simplify H*.H.L$.
4.5 4.5.1
Melody formation Primary derivational rules
The first group of derived melodies is the result of the application of the primary derivational rules, viz. the “opening rule” and the “closing rule”. Their order is immaterial.
(10) Closing Rule Tα =opt: Tα / Tα.
; or simply: Tα =opt: Tα9 / Tα.
L The Closing Rule lowers the second member of the pairs consisting of identical tonal elements, that is it turns the “horizontal” melodic
72
Intonation and Stress: Evidence from Hungarian
stretches into gently descending ones and thus creates new, derived contours out of the basic ones:6
(11) a. H*.L.L$ =opt: H*.L.L9$ (fully falling contour; 1x) b. H*.H$ =opt: H*H9$ (descending contour; 5x) c. H*.H.L$ =opt: H*.H9.L$ (descending–falling contour; )x) The Opening Rule (12) is the opposite of (10). It raises the second member in a sequence of two identical tones and so it turns a horizontal stretch of melody into a gently rising one:
(12) Opening Rule Tα =opt: Tα / Tα.
; or simply: Tα =opt: Tα8 / Tα.
H The newly derived melodies are: 7
(13) a. H*.L.L$ =opt: H*.L.L8$ (falling–rising contour; 7x) b. H*.H$ =opt: H*H8$ (rising contour; 2x) c. H*.H.L$ =opt: H*.H8.L$ (rising–falling contour; 6x) However, the representations of the new contours do not remain in this form, because after the primary derivation has taken place, a simplifying rule that will be called the LH-Rule (14) will modify them:
(14) LH-Rule Tα.Tα =obl: L.H; or simply: Tα.Tα8 =obl: L.H H This simplification is possible because the end-points of the upward pitch movements can be at different heights but this variation does not seem to produce different intonation patterns in Hungarian. Applying (14) to the representations in (13), we get (15):
(15) a . H*.L.L8$ =obl: H*.L.H$ (falling–rising contour; 7x) b. H*.H8$ =obl: L*H$ (rising contour; 2x) c. H*.H8.L$ =obl: L*.H.L$ (rising–falling contour; 6x)
An Autosegmental Analysis of Hungarian Intonation
73
The application of Rule (14) has a further consequence as well: the raised high tone (H8) becomes superfluous and so does the topmost level in (5). The new situation is shown in (16):
(16) Tones and the corresponding pitch levels (Revised version) H H9 = L8 L L9 4.5.2
Secondary derivational rules
After the first phase of melody formation and the simplifications have taken place, the second phase of melody formation begins – the application of the secondary derivational rules. There are three secondary derivational rules; they affect a single melody each, and their order is immaterial. The first one can be called the “Exclamation Rule” (17).
(17) Exclamation Rule a. L =opt: L / H*.
$; or simply: L =opt: L8 / H*.
$
H b. L =obl: H / H*. H
$; or simply: L8 =obl: H9 / H*.
$
L
The result of (17)a, H*.L8$, is uninterpretable because it violates the condition that a raised or lowered tone is only possible if the corresponding non-raised or non-lowered tone is also present in the same representation (cf. section 4.3). Therefore (17)b automatically changes it into the phonetically equivalent, but interpretable representation: H*.H9$. The Exclamation Rule produces a second-type descent out of the half fall, see (18).
(18) H*.L$ =opt: H*.L8$ =obl: H*.H9$ (second-type descending contour; 5x) The next rule (19) also affects the half fall and will be called the “Stylizing Rule” because it produces the stylized fall.
74
Intonation and Stress: Evidence from Hungarian
(19) Stylizing Rule a. L =opt: L+ / H*.
$; or simply: L =opt: L8+ / H*.
$
$; or simply: L8 =obl: H9 + / *.
$
H b. L + =obl: H + / H*. H
L
(19)a raises the second tone and simultaneously furnishes it with the stylizer ( + ). Since the output of (19)a, H*.L8 + $, is uninterpretable, it will automatically change into the phonetically equivalent but interpretable H*.H9 + $; see (20):
(20) H*.L$ =opt: H*.L8+$ =obl: H*.H9+$ (stylized falling contour; 4-x)
The stylizer is not a melodic element, its function is to prevent the second tone from becoming the end-point of a gradual descent that would otherwise develop between the two tones, see (42) below. The insertion of the stylizer is necessary because the stylized fall consists of two horizontal terraces (see section 3.2.7) and these terraces are sharply separated from each other even within a single syllable. The stylizer is the device that keeps the two terraces distinct, see (21)a and (21)b. In the diagram for (21)b the letter á has been doubled, to indicate lengthening:
(21) a. H
b. H
L
L Pál!
H*.H9$ (descent)
Páál!
H*.H9+$ (stylized fall)
The last melody-forming rule (22) will be called the “Appendix Rule” because it produces an appended contour out of a full fall:
An Autosegmental Analysis of Hungarian Intonation
(22) Appendix Rule H*. =opt: 0 /
L. L$; or simply: H*. =opt: 0 /
75
L.L9$
L Rule (22) “decapitates” the fully falling contour by eliminating the big initial drop which is characteristic of the full fall, and turns it into an appended contour:
(23) H*.L.L9$ =opt: L.L9$
4.6
8
Summary of the rules and representations
Before going on to the association conventions, we summarize the rules and representations we have introduced.
(24) (i) 1. 2.
Summary of the melody formation rules Primary derivational rules Closing rule: Tα =opt: Tα9 / Tα. Opening rule: Tα =opt: Tα8 / Tα.
(ii) Simplifying rules 3. Identical tone reduction rule: .Tα =obl: 0/Tα 4. LH-rule: Tα.Tα8 =obl: L.H (iii) Secondary derivational rules $ 5. Exclamation rule: (a) L =opt: L8 / H*. (b) L8 =obl: H9 / H*. $ 6. Stylizing rule: (a) L =opt: L8 + / H*. $ (b) L8 + =obl: H9 + / H*. $ 7. Appendix rule: H*. =opt: 0 / L.L9$ (25) Summary of the representations Full fall Half fall Fall–rise Rise High monotone Descent/ 2nd-type descent
H*.L.L9$ H*.L$ H*.L.H$ L*.H$ H*$ 9 H*.H9$
$
76
Intonation and Stress: Evidence from Hungarian
Rise–fall Monotone-fall Descent-fall Stylized fall Appended contour Preparatory contour
4.7
L*.H.L$ H*.H.L$ H*.H9.L$ H*.H9+$ 10 L.L9$ T (unspecified, but none of the others, typically L)
The association conventions
Since only melodic elements (i.e. tones) can be associated with the syllables of the syllabic tier, the contour end ($) and the stylizer ( + ) are not associated. The melodic elements to be associated with the syllables are the tones H*, H, H9, H9 +, L*, L, L8, and L9. On the syllabic tier, the strings of syllables which carry intonation contours are the contour carriers. Character contours are associated with contour carriers whose initial syllable is primary stressed. Our task here is to show how the elements of any intonation contour can be associated with the syllables of a contour carrier. We need the following association conventions:
(26) Association conventions (ACs) AC 1 (a) The 1st tone is to be associated with the 1st syllable of the contour carrier. (b) The 2nd tone is to be associated with the 2nd syllable of the contour carrier. (c) The 3rd tone is to be associated with the last syllable of the contour carrier. AC 2 If an associable tone remains unassociated after satisfying AC 1, it is to be tied to the syllable associated with the immediately preceding tone within the same contour carrier. AC 3 Each tone spreads to the right, and its spread continues as long as it finds an unassociated syllable within the contour carrier. Let us consider a few examples with one-syllable, two-syllable and more-than-two-syllable contour carriers. The words in the examples are Pál ‘Paul’, Mari ‘Mary’, informal, Angéla néni ‘Aunt Angela’ and és ‘and’.
An Autosegmental Analysis of Hungarian Intonation
(27) Full fall a. | 1Pál. |
b. | 1Mari. |
σ
σ
σ
H*.L.L9$ (AC 1a)
H*.L.L9$ (AC 1c)
H*.L.L9$ (AC 2)
σ
σ
σ
σ
H*.L.L9$ (AC 1a)
σ
H*.L.L9$ (AC 1b)
σ
H*.L.L9$ (AC 1c)
c. | 1Angéla néni. |
σσσσσ
σσσσσ
σσσσσ
σσσσσ
H*.L.L9$ (AC 1a)
H*.L.L9$ (AC 1b)
H*.L.L9$ (AC 1c)
H*.L.L9$ (AC 3)
(28) Rise a. | (És) 2Pál? |
b. | (És) 2Mari? |
σ
σ
L*.H$ (AC 1a)
L*.H$ (AC 2)
σ σ
σ σ
L*.H$ (AC 1a)
L*.H$ (AC 1b)
c. | (És) 2Angéla néni? |
σσσσσ
σσσσσ
σ σ σ σ σ11
L*.H$ (AC 1a)
L*.H$ (AC 1b)
L*.H$ (AC 3)
77
78
Intonation and Stress: Evidence from Hungarian
(29) Descent a. | (És) 5Pál? |
b. | (És) 5Mari? |
σ
σ
H*.H9$ (AC 1a)
H*.H9$ (AC 2)
σ σ
σ σ
H*.H9$ (AC 1a)
H*.H9$ (AC 1b)
c. | (És) 5Angéla néni? |
σ σ σ σ σ
σ σ σ σ σ
σ σ σ σ σ12
H*.H9$ (AC 1a)
H*.H9$ (AC 1b)
H*.H9$ (AC 3)
(30) Rise–fall a. | 6Pál? |
b. | 6Mari? |
σ
σ
σ
L*.H.L$ (AC 1a)
L*.H.L$ (AC 1c)
L*.H.L$ (AC 2)
σ σ
σ σ
σ σ
L*.H.L$ (AC 1a)
L*.H.L$ (AC 1b)
L*.H.L$ (AC 1c)
c. | 6Angéla néni? |
σσσσσ
σσσσσ
σσσσσ
σ σ σ σ σ13
L*.H.L$ (AC 1a)
L*.H.L$ (AC 1b)
L*.H.L$ (AC 1c)
L*.H.L$ (AC 3)
An Autosegmental Analysis of Hungarian Intonation
(31) Stylized fall a. | 4-Pál! |
b. | 4-Mari! |
σ
σ
H*.H9+$ (AC 1a)
H*.H9+$ (AC 2)
σ σ
σ σ
H*.H9+$ (AC 1a)
H*.H9+$ (AC 1b)
c. | 4-Angéla néni! |
σσσσσ
σσσσσ
σ σ σ σ σ14
H*.H9+$ (AC 1a)
H*.H9+$ (AC 1b)
H*.H9+$ (AC 3)
(32) Appended contour a. (| 6Maga az,) | 3 Pál? |
b. (| 6Maga az,) | 3 Mari? |
σ
σ
L.L9$ (AC 1a)
L.L9$ (AC 2)
σ σ
σ σ
L.L9$ (AC 1a)
L.L9$ (AC 1b)
c. (| 6Maga az,) | 3 Angéla néni? |
σσσσσ
σσσσσ
σ σ σ σ σ15
L.L9$ (AC 1a)
L.L9$ (AC 1b)
L.L9$ (AC 3)
(33) Preparatory contour | 1Gyere! |
Ide
1ne firkálj! |
| σ σ |
σ σ
σ σ σ |
T
H*.L.L9$
H*.L.L9$
79
80
4.8
Intonation and Stress: Evidence from Hungarian
The rearrangement rules
In certain cases the association lines may be or must be rearranged. Rearrangement is obligatory in the characters H*.H9$, H*.H9.L$, H*.H9 + $ and in the appended contour L.L9$. Each of these melodies contains a sequence of T α.Tα9. In these contours the Tα9 tone is associated with the 2nd syllable and from that syllable it spreads to the right. Rearrangement here takes place in such a way that all the association lines between the Tα9 tone and the syllables, except for the rightmost one, are cut off one by one, proceeding from left to right, while the preceding tone (Tα ) spreads onto the syllables which have thus become toneless; see (34).
(34) Rearrangement of association lines in descending sequences (Obligatory)
σ σ σ
σ σ σ =
=obl: ( )
( )
Tα * .Tα9
Tα * .Tα9
Accordingly, (35)a: | [És] 5Angéla néni? | becomes (35)b; (36)a: | )Angéla néni? | becomes (36)b; (37)a: | 4-Angéla néni! | becomes (37)b; and (38)a: | 3Angéla néni | becomes (38)b, obligatorily:
(35) a. = (29)c
σ σ σ σ σ
=obl:
b.
H*.H9$ (36) a. σ σ
σ σ σ
H*.H9$ =obl:
H*.H9.L$ (37) a. = (31)c
b. σ
σ σ σ σ
H*.H9.L$
σ σ σ σ σ
=obl:
b. σ
H*.H9+$ (38) a. = (32)c
σ σ σ σ σ
σ σ σ σ σ L.L9$
σ σ σ σ
H*.H9+$ =obl:
b.
σ σ σ σ σ L.L9$
Rearrangement of the association lines is optional in the contours L*.H$ and L*.H.L$, that is in rising sequences. In these melodies the tone H is associated with the 2nd syllable and from that syllable it spreads to the
An Autosegmental Analysis of Hungarian Intonation
81
right. Rearrangement here means that all the association lines between the H tone and the syllables, apart from the rightmost one, are cut off one by one, proceeding from left to right, while the preceding tone (L*) spreads onto the syllables which have thus become toneless; see (39).
(39) Rearrangement of association lines in rising sequences (Optional)
σ σ σ
σ σ σ =
=opt: L*.H
L*.H
Rule (39) enables the speaker to replace (40)a: | [És] 2Angéla néni? | by (40)b, or to replace (41)a: | ˆAngéla néni? | by (41)b, although the change is not obligatory.
(40) a. = (28)c
σ σ σ σ σ
=opt:
b. σ σ σ σ σ
L*.H$ (41) a. = (30)c
σ σ σ σ σ L*.H.L$
L*.H$ =opt:
b.
σ σ σ σ σ L*.H.L$
Applying the Association Conventions (26) and the Rearrangement Rules (34) and (39) always results in associations which satisfy the wellformedness conditions set up by Goldsmith (1976, 1990: 319).
4.9
Phonetic realization
In contours where a particular tone spreads to the right, the spreading tone is often realized not as a monotone but as a gently rising or gently descending stretch of melody, under the influence of the following tone. These possibilities are captured by an obligatory and an optional phonetic realization rule: see (42) and (45).
(42) Rule of Gradual Descent (Obligatory) If, in a contour, tone Tα is followed by Tα9 and Tα spreads to the right, then the syllables associated with Tα will gradually descend towards the pitch height of the syllable associated with Tα9, unless Tα9 is furnished with the stylizer (i.e. it is Tα9+).
Intonation and Stress: Evidence from Hungarian
82
Thus the character contours H*.L.L9$, H*.H9$, H*.H9.L$, and the appended contour L.L9$ are realized on the string Angéla néni (‘Aunt Angela’), as shown in (43):
(43)
a. σ σ σ σ σ
b. σ σ σ σ σ
H*.L.L9$
H*.H9$
c. σ σ σ σ σ H*.H9.L$
d. σ σ σ σ σ 16
L.L9$
But the stylized fall (H*.H9 + $) is exempted from the gradual descent because of the presence of the stylizer on the second tone. It is realized as a higher terrace abruptly followed by a lower one (with lenghtening of the final syllable as described in section 3.2.7), see the realization of |4-Angéla néni! | in (44):
(44) = (37)b
σ
σ σσσ
H*.H9+$ The other phonetic realization rule concerns rises and is optional:
(45) Rule of Gradual Rise (Optional) If, in a contour, tone L is followed by tone H and tone L spreads to the right, the syllables associated with L may gradually rise towards the pitch height of the syllable associated with H.
This means that, if we apply (45), the contours H*.L.H$, L*.H$ and L*.H.L$ can be realized as shown in (46). The carrier phrase in (46) and in the rest of the examples is Angéla néni.17
An Autosegmental Analysis of Hungarian Intonation
83
(46)
a. σ σ σ σ σ H*.L.H$
b. σ σ σ σ σ L*.H$
c. σ σ σ σ σ L*.H.L$
But if we do not apply (45), the melodies will have the following shapes:
(47)
a. σ σ σ σ σ H*.L.H$
b. σ σ σ σ σ L*.H$
c. σ σ σ σ σ L*.H.L$
While in (46) the rise is gradual, in (47) it is abrupt and restricted to the final portion of the rising stretch (“end-rising”). But the phonetic realization of the contours L*.H$ and L*.H.L$ can be of a third kind as well. This is achieved when the speaker does not carry out (39), the optional rearrangement of association lines in LH sequences. In this case the rise is abrupt and takes place in the initial part of the contour (“front-rising”):18
(48)
a. σ σ σ σ σ L*.H$
b. σ σ σ σ σ L*.H.L$
The “front-rising” variants of the rising and rising–falling contours (48)a, (48)b are never obligatory but their likelihood seems to increase as the number of syllables in the carrier phrase increases.19
84
Intonation and Stress: Evidence from Hungarian
Variations like the “gradual rising” (46), “end-rising” (47) and “frontrising” (48) show that the phonetic realization of the LH sequence is not simply a matter of interpolation between an L and an H tone. The preparatory contour has an unspecified tonal representation (T). This is realized most typically as a monotone on level L or as a gently rising or gently descending contour, no part of which is lower than level L (see section 3.5). The example in (49) is | De ezek ˆkész vannak? | ‘But are these finished?’, where the words de ‘but’ and ezek ‘these’ carry the preparatory contour.
(49)
σ σσ
σ σσ
T
L*H.L$
Before finishing this section, it is appropriate to point out that the exact correspondence between micromelodic phonetic realizations and our autosegmental representations requires a lot of further research, and our apparatus may eventually have to be refined. For example, it is possible to find cases where a fall or a descent has its contour peak not at the very beginning but in the second half or towards the end of the initial syllable, with which the pitch accent is associated. In fact, I have sometimes heard the peak shifted as far as to the second syllable in second-type descents, as in (50): | 5Milyen szépen énekel! | ‘How beautifully (s)he sings!’
(50)
σ σ σ σ σ σ σ H*.H9$ Although the starred tone here is associated with the first syllable, it is actually aligned with the second. Following Ladd (1983), we can call this kind of variation [ + delayed peak].20
An Autosegmental Analysis of Hungarian Intonation
4.10
85
Structure of the Hungarian intonation phrase
All the legitimate contour combinations (tunes) within Hungarian IPs (as defined in section 3.6.1) are generated by the finite-state grammar shown in (51):
(51) Structure of the Hungarian IP Preparatory
T |
Scale
Terminal
H*.L$
H*.L.L9 H*.L$ H*.L.H$ L*.H$ H*$ H*.H9$ L*.H.L$ H*.H.L$ H*.H9.L$ H*.H9+$ L.L9$
|
If the IP has a half-falling terminal (i.e. the contour H*.L$), it must end in an audible pause, in all other cases the final audible pause is possible but not obligatory. Consequently, a half fall terminal can only come into being in that step of the melodic segmentation of the utterance in which the pauses are assigned (see section 1.2.2). While the character contours are expansions of an initial starred tone, the appended contour has no initial starred tone. In this framework the contour occupying the terminal part is not associated with an independently variable phrase accent and boundary tone in the way it would be in Pierrehumbert’s framework (Pierrehumbert 1980: 29). This follows from our viewing the Hungarian terminal contours as meaningful wholes (see sections 1.2.1 and 3.2.1). In addition, these elements of the Pierrehumbert framework are problematic for the Hungarian rise–fall (our L*.H.L$). In an adaptation of that framework, the Hungarian rise– fall is represented as L*HL%, with H being a phrase accent and the final L a boundary tone (cf. Ladd 1996: 116, 133, 136; or Grice, Ladd and Arvaniti 2000; specifically about this Hungarian contour). However, in Pierrehumbert’s original framework (see Ladd 1996: 88) the final L% boundary tone after an H phrase accent indicates a level sustention of the previous tone, and so its ability to express the final fall of the Hungarian rise–fall is doubtful.21
86
Intonation and Stress: Evidence from Hungarian
4.11
Summary
In this chapter the Hungarian intonation contours were submitted to an autosegmental phonological analysis. This analysis is deeper than the surface-phonological analysis given in Chapter 3 because, by breaking up the surface contours into tonal constituents, it derives the twelve surface contours from three basic ones and explicitly shows the genetic relations (as well as the similarities and differences) between the surface contours.
5 The Melodic Segmentation of Hungarian Utterances
5.1
Introduction
In Chapter 5 we attempt to establish rules for the melodic segmentation of Hungarian utterances, that is for the process whereby the intonation contours that linearly follow one another in the spoken Hungarian sentence are formed and the intonational phrases established.1 At the same time, we explore how the intonation contours depend on one another in the spoken sentence and how they are related to surface syntactic structure. When the spoken realization of a Hungarian sentence contains more than one intonation contour, these contours depend on each other in certain ways, and these dependencies can be expressed in rules. If there were no such dependency rules, we would not be able to explain why the melodic structure of (1)a is correct while the melodic structure of (1)b is not.
(1) A
tengeralattjáró
the submarine
legénysége
szabadságra
ment.
crew-its
holiday-on
went-3sg
‘The crew of the submarine went on holiday.’ a. | A [tengeralattjáró ´legénysége | `szabadságra ment. | sége sza ten legény A geralattjáró badság ra ment. b. *| A ´tengeralattjáró | [legénysége `szabadságra ment. |
A
tenger
alattjá
ró le
génysége szabad ságra 87
ment.
88
Intonation and Stress: Evidence from Hungarian
In order to identify these dependencies, we first have to clarify the notions of sentence and utterance. This will be done in section 5.2, where the notion of the tonosyntactic block will also be introduced. In section 5.3 we deal with the major tonosyntactic blocks: these are melodically autonomous or semi-autonomous blocks within utterances. The central part of the chapter is section 5.4: this is a detailed analysis of minor tonosyntactic blocks, which can be equivalent or complementary, or both equivalent and complementary. Their common characteristic is that each type occurs in a series of non-final block(s) and final block, and the final (rightmost) block melodically constrains the non-final one(s); that is there is a right-to-left melodic dependency between the final and non-final blocks. At the end of the chapter, in section 5.5, we discuss some deviations from the intonation predicted by our rules of melodic segmentation. Some of the ideas in this chapter were first presented in Varga (1985, 1989, 1993, 1994).
5.2
Sentence and utterance
It will be recalled that according to the most widely known Hungarian syntactic model (É. Kiss 1981, 1987, 1992), surveyed in Chapter 1, section 1.2.3, the Hungarian sentence consists of a topic (T) and a comment (C); the latter has four syntactic positions: (a) the Q or quantifier position; (b) the F or focus position; (c) the V or verb position; and (d) the Pv or postverbal position. All these syntactic positions are illustrated in (2).
(2)
János mindig
kávét
John
coffee-ACC drinks the snackbar-in
always Q
T
iszik
F
a
büfében.
V
Pv
C Sentence
‘John always drinks coffee in the snackbar.’ | [János [mindig [kávét iszik a `büfében. | Já min nos dig kávét iszik a büfé ben.
2
The Melodic Segmentation of Hungarian Utterances
89
Any string of Hungarian words that realizes the structure shown in (2) is a Hungarian sentence from a structural point of view, even if the constituents carrying given information are ellipted. For instance, in answer to the question Mit iszik János mindig a büfében? ‘What does John always drink in the snackbar?’, we can simply say the “elliptical sentence” (3):
(3)
0
0
Kávét
0
0
V
Pv
‘Coffee.’
coffee-ACC Q
F
T
C Sentence
But expressions such as (2) and (3) are sentences not only from a structural point of view, but also from a functional point of view, because they perform independent illocutionary acts. They are structural-functional sentences. A stretch of language that performs an independent illocutionary act, whether or not it is a sentence structurally, is a functional sentence. The examples in (4) are all functional sentences, although they are not sentences structurally because they do not realize the construction shown in (2) and (3):
(4) a. | `Igen. |
‘Yes.’
b. | `Nem. |
‘No.’
c. | 5Jé! |
‘Well!’
d. | `Persze. |
‘Of course!’
e. | 6Tessék? |
‘Pardon?’
f. | `Jaj! |
‘Ouch!’
g. | [Mi a `fene! |
Approx.: ‘What the hell!’
So sentences are basically of two kinds: they are either structural-functional as in (2) and (3), or only functional as in (4). An utterance (Utt) is not a syntactic but a prosodic unit; it is at the top of the hierarchy of prosodic units. The Utterance, the Intonational Phrase, the Accent Phrase, the Foot and the Syllable are the units of the prosodic hierarchy that we shall distinguish as relevant to the description
90
Intonation and Stress: Evidence from Hungarian
of stress and intonation in Hungarian.3 An utterance can be simple or complex. A simple utterance is the spoken realization of a highest-ranking sentence (functional or structural-functional). A highest-ranking sentence (HRS) is a sentence which is not itself part of an even higher-ranking sentence.4 For instance, (1)a, (2), (3) and the sentences in (4) are all simple utterances, containing one HRS each. Utterances can also be complex and these will be dealt with in section 5.3. We shall use the term tonosyntactic block to refer to syntactic units that are relevant to establishing the melodic structure of utterances.5 Thus the highest-ranking sentence (whether structural-functional or only functional) is obviously a tonosyntactic block because its spoken realization always begins and ends with an IP boundary. The melodic segmentation of an utterance is the identification of the tonosyntactic blocks inside the utterance, with the assignment of a melodic prosodeme (i.e. an intonation contour) to each of these blocks, and it ultimately results in the establishment of intonational phrases. In the analyses, the labels of tonosyntactic blocks will be put between braces, e.g. {HRS}. (5) is the analysis of our former example (1)a. Being coextensive with an HRS, it is a simple utterance, which consists of two intonational phrases:
(5) | A [tengeralattjáró ´legénysége | `szabadságra ment. | IP
IP Utt = {HRS}
As was shown in section 3.6, the division of the Hungarian utterance into IPs is in accordance with the Strict Layer Hypothesis, that is utterances are divided into IPs exhaustively and no IP can contain another IP (see Selkirk 1984: 26).
5.3 5.3.1
Major tonosyntactic blocks Utterances with more than one highest-ranking sentence
A complex utterance is an utterance which consists of more than one major tonosyntactic block. The major tonosyntactic blocks are: the highestranking sentence (HRS), the non-quoting inorganic block (NQIO) and the quoting inorganic block (QIO). The latter two will be discussed in section 5.3.2.
The Melodic Segmentation of Hungarian Utterances
91
The HRS is a melodically autonomous tonosyntactic block because its last (rightmost) character contour is not determined or constrained by any other character contour within the utterance; this contour is chosen freely by the speaker to express the communicative type (direct illocution) of that particular HRS and the special attitude that (s)he associates with that HRS. When there are several HRSs in a complex utterance, each of them is associated with the performance of a separate illocutionary act, but one of these acts is usually more important than the others. For instance, in the complex utterance shown in (6), in which a functional HRS (hurrá ‘hurray’) is followed by a structural-functional HRS (indulunk ‘start-1pl’), the latter HRS is the semantically more important one, it is the “nucleus” of the utterance.
(6) Hurrá, hurray
indulunk! start-lpl
‘Hurray, we are starting out!’ | `Hurrá, | `indulunk! | IP={HRS} IP={HRS} Utt
Nevertheless, even the “less important” HRS within the utterance (hurrá) is an autonomous tonosyntactic block: its final character contour is not determined or constrained by any other character contour within the utterance. Although the HRSs in complex utterances are in separate IPs, these IPs are prosodically united into an utterance. Prosodic unity of the IPs belonging to the same utterance can be achieved by various means. One of these is having less substantial pauses between parts of the same utterance than between separate utterances; in (6), for instance, even if there is a pause between the HRSs hurrá and indulunk, it must be relatively short to prevent interpretation of the two HRSs as two separate utterances. Another device of utterance formation is the realization of inter-IP downdrift over the parts of the utterance (see section 3.7). For instance, in (6) the second full fall has a slightly lower peak than the first. In (7) we find some further examples of complex utterances consisting of more than one HRS, each taking a separate IP.
92
Intonation and Stress: Evidence from Hungarian
(7) a. Késo late
van,
nem
érted?
is
not
understand-2sg
‘It’s late, don’t you understand?’ | `Késo van, | 6nem érted? | b. Indulunk, start-lpl
jó? good
‘We are starting out, OK?’ | `Indulunk, | 6jó? | c. Bejöhetsz, in-come-can-2sg
de
eddig
hol
voltál?
but
so-far
where
were-2sg
‘You may come in, but where have you been until now?’ | 7Bejöhetsz, | de eddig `hol voltál? | d. Mit what-ACC
gondolsz,
ott
vannak?
think-2sg
there
are-3pl
‘Do you think they are there?’ (Literally: *‘What do you think, are they there?’) | `Mit gondolsz, | 6ott vannak? | e. Fordítva the-other-way-round
én
mondtam!
I
said-1sg
‘The other way around, it was me who said it!’ | `Fordítva, | `én mondtam! | f. Angéla, Angela
kérsz
kávét?
want-2sg
coffee-ACC
‘Angela, would you like some coffee?’ | `Angéla, | 6kérsz kávét? | g. Kérsz
kávét,
Angéla?
‘Would you like some coffee, Angela?’ | 6Kérsz kávét, | `Angéla? | In the complex utterances of (8) below, an HRS is inserted into the body of another (more important) HRS and thus cuts the latter in two. The insertions form separate intonational phrases; moreover, their insertion
The Melodic Segmentation of Hungarian Utterances
93
causes the preceding part and the subsequent part of the matrix HRS to become IPs too. They are normally separated from the rest of the utterance by pauses. 6 (8)
a.
Találkoztam –
jól
ülsz? –
az
Angélával!
met-1sg
well
sit-2sg
the
Angela-with
‘I’ve met – are you sitting comfortably? – Angela.’ | [Találkoztam | - – 6jól ülsz? – | - az `Angélával! | IP
IP = {HRS}
IP
Utt = {HRS} b. Összesen mennyit?,
tán
két évet
volt
ott.
all-in-all how-much-ACC perhaps two year-ACC was-3sg there ‘All in all, he spent – how long? – perhaps two years there.’ | [Összesen | - 6mennyit?, | - tán `két évet volt ott. | c.
5.3.2
Elkezdtem, de tudod, hogy van?, nem volt idom. started-1sg but know-2sg how is not was-3sg time-my ‘I started it but – you know how it is – I had no time.’ | [Elkezdtem, de | - 7tudod, hogy van?, | - [nem volt `idom. |
Utterances with inorganic blocks
Complex utterances may also contain inorganic blocks; these are not organic parts of utterances and lack a major stress. They correspond to Bing’s “class 0 expressions” (1980: 21–36) and have the following main types:
(9) Main types of inorganic blocks a. Sentence qualifiers : these are adverbial expressions conveying the speaker’s comment on the form or purpose of what (s)he is saying, or on the conditions under which (s)he is speaking, e.g. oszintén szólva ‘frankly speaking’; az igazat megvallva ‘telling you the truth’; jobban mondva ‘putting it better’; hogy finom legyek ‘to be polite’; hogy oszinte legyek ‘to be frank’; hogy rövid legyek ‘to be brief’; hogy pontos legyek ‘to be precise’; ha oszinte lehetek ‘if I may be frank’; ha szabad így fogalmaznom ‘if I may put it this way’; ha jól emlékszem ‘if I remember right’; akarom mondani ‘I mean’; etc.
94
Intonation and Stress: Evidence from Hungarian
b. Vocatives: e.g. kedvesem ‘darling’; doktor úr ‘doctor’; te orült ‘you c.
fool’; etc. Expletive verb phrases directed at the hearer: e.g. kérlek (szépen) ‘please’ (literally: ‘I ask you (nicely)’) (“tu” form), kérem (szépen) ‘please’ (“vous” form); tudod ‘you know’ (“tu” form); tudja ‘you know’ (“vous” form); látod ‘you see’ (“tu” form); látja ‘you see’ (“vous” form); nézd ‘look’ (“tu” form); nézze ‘look’ (“vous” form); mondd ‘say’ (“tu” form), mondja ‘say’ (“vous” form);
csókolom literally: ‘I kiss you’; etc. d. Quoting clauses: e.g. kérdeztem ‘I asked’; mondta Éva ‘said Eve’; etc. Since the prosodic behaviour of the first three kinds of inorganic blocks (9)a, (9)b and (9)c is identical, we shall group them together under the name of non-quoting inorganic blocks (NQIO). At the beginning of an utterance they carry a preparatory contour (10)a, (10)b. In utterancemedial position they form part of the contour that has started immediately before them (11)a, (11)b.
(10) a. Látod, see-2sg
megint
nem
figyelsz.
again
not
listen-2sg
‘You see, you are not listening again.’ | Látod, `megint
nem
{NQIO}
figyelsz. |
{HRS} Utt = IP
b. Angéla, Angela
kérsz
kávét?
want-2sg
coffee-ACC
‘Angela, would you like some coffee?’ | Angéla, 6kérsz kávét? | {NQIO}
{HRS}
Utt = IP (11)
a. Elhoztam,
hogy
brought-1sg that=CONJ
a
tárgyra
térjek
a
pénzt.
the subject-to turn-IMP-1sg the money-ACC
The Melodic Segmentation of Hungarian Utterances
95
‘I’ve brought – to come to the point – the money.’ | [Elhoztam, - hogy a
]tárgyra térjek, | - a `pénzt. | {NQIO}
IP
IP Utt = {HRS}
b.
Kérsz,
Angéla,
kávét?
want-2sg
Angela
coffee-ACC
‘Would you like, Angela, some coffee?’ 7
| 6Kérsz, - Angéla, - kávét? | {NQIO} Utt = {HRS}
At the end of the utterance, non-quoting inorganic blocks form a separate appended contour if the final contour of the HRS before them is an end-falling character (12)a, or if it is a full fall or half fall from which they are separated by a pause (12)b. In all other cases they form a continuation of the final contour of the HRS before them. For instance, in (12)c the non-quoting clause continues the rise that began on the HRS. (12)
a.
Kérsz
kávét,
Angéla?
‘Would you like some coffee, Angela?’ | 6Kérsz kávét, | - 3 ]Angéla? | IP={HRS}
IP={NQIO} Utt
b.
Nem tudom,
hogy
oszinte
legyek.
not know-1sg
that=CONJ
frank
be-IMP-1sg
‘I don’t know, to be frank.’ 8
| `Nem tudom, | - hogy 3 ]oszinte legyek. | {HRS}
{NQIO} Utt = IP
96
Intonation and Stress: Evidence from Hungarian
c.
És
ha
kisajátítják,
hogy
finom
legyek?
and
if
expropriate-3pl
that=CONJ
mild
be-IMP-1sg
‘And [what] if they expropriate it, to use a decent word?’ | És ha ´kisajátítják, - hogy ]finom legyek? | {HRS}
{NQIO} Utt = IP
As has been pointed out above, the inorganic blocks can have unstressed and minor-stressed syllables, but no major stresses. If a stretch of language that would be an inorganic block without major stresses occurs with major stresses, it is no longer an inorganic block but either an HRS or a sub-HRS tonosyntactic block. For instance, while the vocative Angéla was inorganic (and unstressed or minor-stressed) in (10)b, (11)b and (12)a; it was a separate HRS (and major-stressed, in a separate IP) in (7)f and (7)g. The fourth kind of inorganic block is the quoting inorganic block (QIO), which is a clause that normally occurs after or inside a direct quotation and lacks a major stress, for example kérdeztem ‘I asked’ or mondta Éva ‘said Eve’. In utterance-medial position it behaves similarly to the nonquoting inorganic block, but in utterance-final position it always forms a separate IP with an appended contour (and not only after the full fall or half fall, or the end-falling contours as the non-quoting inorganic blocks do); see the examples in (13). Since the sentence quoted is already an utterance in itself, the quoting clause added to it must also be regarded as an utterance, and the two together form a so-called Quotation-Carrying Utterance (QuoUtt). This is a compound prosodic domain in the sense of Ladd (1996: 244–6).
(13)
a.
Kérsz
kávét,
Angéla?
want-2sg
coffee-ACC
Angela
–
kérdezte
Péter.
asked-3sg
Peter
‘“Would you like some coffee, Angela?”, asked Peter.’ | 6Kérsz kávét, | - 3 ]Angéla? | - 3 ]kérdezte ]Péter. | IP = {HRS}
IP = {NQIO} Utt
Utt = IP = {QIO} QuoUtt
The Melodic Segmentation of Hungarian Utterances
b.
És
ha
kisajátítják?
and
if
expropriate-3pl
–
kérdezte
Jóska.
asked-3sg
Joe
97
‘And [what] if they expropriate it? – asked Joe.’ | És ha ´kisajátítják? | - 3 ]kérdezte ]Jóska. | Utt = IP = {HRS}
Utt = IP = {QIO} QuoUtt
c.
Elmehetek,
–
leave-can-1sg
szólt
a
kollégám.
spoke-3sg
the
colleague-my
‘I can leave, – said my colleague.’ | 7Elmehetek, | - 3 ]szólt a ]kollégám. | d.
Milyen
szép!
how
beautiful
–
suttogta
Mari.
whispered-3sg
Mary
‘How beautiful, – whispered Mary.’ | 5Milyen szép! | - 3 ]suttogta ] Mari. | e.
Nem
hiszem,
not
believe-1sg
–
mondta
a
miniszter.
said-3sg
the
minister
‘I don’t think so, – said the minister.’ | `Nem hiszem, | - 3 ]mondta a ]miniszter. |
Although the intonational problems of inorganic blocks deserve further study, we cannot pursue this topic any further here. From now on we shall concentrate on the minor tonosyntactic blocks, that is on the tonosyntactic blocks that can be found within highest-ranking sentences, and on the dependency rules between them.
5.4 Minor tonosyntactic blocks: below the highest-ranking sentence 5.4.1
Division into dominant and dependent minor blocks
By minor tonosyntactic blocks or minor blocks (MBs), for short, I mean those major-stressed syntactic constituents that are the immediate or non-immediate constituents of highest-ranking sentences: the sub-HRS tonosyntactic blocks (see section 1.2.2). The input to the division into
98
Intonation and Stress: Evidence from Hungarian
MBs is the stressed syntactic S-structure of the HRS, produced in the stress fixing stage of the syntax–phonology mapping: (14) a.
Nem léphetnek
egyszerre,
összedol a
mert
híd.
not step-can-3pl simultaneously because collapses the bridge ‘They can’t walk in step because the bridge collapses.’ S VP V′ V
Sn′ :
* Nem léphetnek
XP
S
* egyszerre,
* * mert összedol a híd.
Pv
Pv
V
C
In the course of melodic segmentation the minor tonosyntactic blocks (those constituents of the syntactic surface structure that bear pitch accent marks) within the HRS will be demarcated: the structure in (14)a will be divided into two minor blocks, and each of these will in turn be divided again into two minor blocks, as in (14)b: (14)
b.
* Nem léphetnek {MBdep}
* egyszerre, {MBdom}
* összedol
mert
{MBdep}
{MBdep}
a
* híd.
{MBdom}
{MBdom} {MBdom} = {HRS}
When a syntactic constituent (other than an inorganic block) occurs without a major stress (pitch accent mark), then it does not constitute a separate MB. For instance, in (14)b, the unaccented verb léphetnek after the negative particle nem becomes part of the MB starting with the
The Melodic Segmentation of Hungarian Utterances
99
negative particle, whereas the unaccented words mert and a become part of the MBs with összedol and híd, respectively. Each block (HRS or MB) which is composed of MBs is the matrix block of those MBs. The rightmost character contour in a matrix block is the dominant contour in that matrix block, and the MB containing the dominant contour is the dominant MB. The dominant MB (MBdom) is preceded by its sister(s): none, or one, or more than one dependent MB (MB dep) within the matrix block. The rightmost character contour in a dependent MB is a dependent contour (and at the same time dominant in relation to the melody of its own internal sister). The choice of a dependent contour is not free because it is constrained by the dominant contour of the matrix block. 5.4.2
Assignment of melodies
The next step is to supply the rightmost major-stressed syllable of the highest dominant MB of the HRS with the character contour which is required by the communicative type (direct illocution) and special attitude of the HRS. This rightmost melody constrains the melodic possibilities before it in the HRS. In each series of sister MBs the rightmost character contour of the rightmost MB determines what particular character contour will be assigned to the rightmost major-stressed syllable in each of the sister MBs; or else it leaves these syllables without a melody. This process starts at the top and is repeated at each lower level until reaching the lowest MBs. After the melodies have been assigned, all major-stressed syllables that remain without a character contour will automatically receive a half-falling contour, that being the default melody of major stresses. Our melody rules will have to define what particular contours can or must be assigned to the major-stressed syllables of the dependent blocks and under what circumstances. Whenever the contour assigned is other than the half fall, the contour is a terminal one and will end in an IP boundary. When the melodies have been assigned by the melody rules, certain kinds of contour replacement can take place (see section 5.4.7). Finally, pauses may be assigned. These do not affect the IP boundaries that have already emerged (at the end of each character contour other than half-falling), but they introduce new IP boundaries after half-falling characters.9 The result of melodic segmentation is a fully intonated syntactic surface structure (Sn″), which at the same time is the underlying sentence-level phonological representation (P1); that is Sn″ = P1 (Selkirk 1984: 34). Example (15) displays the steps in the melodic segmentation of (14). (In order to save space, the MB structure is no longer shown.)
100
Intonation and Stress: Evidence from Hungarian
(15)
Melodic segmentation a.
Input: Sn′ * Nem
b.
mert
* összedol
a
* híd.
* egyszerre,
léphetnek
mert
* összedol
* a `híd. |
Establishment of the rightmost contour of the highest dependent MB: * | Nem
d.
* egyszerre,
Establishment of the dominant contour of the HRS: * | Nem
c.
léphetnek
léphetnek
´egyszerre, | mert 8
* összedol
a `híd. |
Establishment of the rightmost contours of the dependent MBs at the next lower level: [Nem
léphetnek
´egyszerre, | mert
[összedol
a `híd. |
NB: Half falls are assigned to all remaining pitch accents by default. e.
Assignment of pauses: | [Nem léphetnek
f.
´egyszerre, | - mert
[összedol
a `híd. | -
´egyszerre, | - mert
[összedol
a `híd. | -
Output : Sn″ = P1: | [Nem léphetnek
The ultimate domain of Hungarian speech melody is the autonomous block, typically the highest-ranking sentence. It has often been claimed that in English any contour can be used over any stretch of the utterance as long as it expresses the speaker’s attitude (see Pike 1945: 23), or the speaker’s assessment of the stretch as being new or given information for the hearer (see Brazil, Coulthard and Johns 1980: 13–19; Brazil 1985). This principle of “any contour anywhere” has been convincingly criticized by Bing (1980), who claims that the ultimate domain of intonation is the sentence. Her criticism is now confirmed by our findings in Hungarian. 5.4.3
Equivalent blocks (EBs)
On the basis of two possible kinds of relationship between the MBs of a matrix block, MBs can be classified into (a) equivalent blocks; (b) complementary blocks; and (c) blocks that allow both interpretations. Equivalent blocks are sister MBs between which there is an equivalence relationship, that is they are expressions that have to be regarded as existentially equivalent. The aim of the second EB is not to continue or
The Melodic Segmentation of Hungarian Utterances
101
complete the first; the two are in some sense “the same”. This “sameness” is expressed melodically: there is right-to-left contour concord between the two EBs. The rightmost character contour of the non-final EB (i.e. the dependent EB) becomes identical with the rightmost character contour of the final EB (i.e. the dominant EB). Contour concord is ensured by the melody rule in (16). In (16), as well as in the other melody rules, we shall observe the following formalization conventions. Minor stresses will be ignored as they do not affect the shape of the intonation contours. The letter s will represent a string of syllables (consisting of a minimum of one syllable) which carries a melodic prosodeme. Round brackets will be used in the following way: (s) means optional syllable strings, (s)1 means repeatable syllable strings, and (s)0 means syllable strings that are optional but also repeatable. The letters nf and f represent non-final and final (i.e. dependent and dominant) blocks, respectively. ST stands for the “syntactic type” of the non-final and final blocks. The letter s with an asterisk (*) over it means a string of syllables beginning with a major-stressed (i.e. pitch accented) syllable.
(16) EB Melody Rule (i): * ω s =obl: s | / [nf where
* (s) (s)o
]
[f (Y)
ω
s|]
ω
s = any character contour, equivalent in nf and f.
Y = material of the final EB up to the dominant contour. Condition: nf and f belong to ST1. ST1: nf = clausal or non-clausal constituent preposed from F position f = main clause or comment whose F position contains a demonstrative referring to the preposed clausal or non-clausal constituent. In the following examples the non-final EBs will be italicized. In (17) the subclause hogy elkésett is preposed from behind the demonstrative az, together with which it occupies the F position of the main clause.
(17) a.
Hogy
elkésett,
az
bosszantotta.
that=CONJ
was-late-3sg
that=DEM
annoyed-3sg
‘It was the fact that he was late, that annoyed him.’ (Literally: ‘That he was late, that annoyed him.’) | Hogy `elkésett, | `az bosszantotta. |
102
Intonation and Stress: Evidence from Hungarian
b. És
elkésett,
hogyha
and
az
bosszantotta?
if
‘And [what] if it was the fact that he was late that annoyed him?’ | És hogyha ´elkésett, | ´az bosszantotta. | c. Hogy
elkésett,
az
bosszantotta?
‘Was it the fact that he was late that annoyed him?’ | Hogy 6elkésett, | 6az bosszantotta? | In (18) the element preposed from F position is the non-clausal constituent a telefon:
(18) a.
A
telefon,
az
szól.
the
telephone
that=DEM
rings
‘It is the telephone that is ringing.’ (Literally: ‘The telephone, that’s ringing.’) | A `telefon, | `az szól. | b. És and
ha
a
telefon,
az
szól?
if
‘And [what] if it is the telephone that is ringing?’ | És ha a ´telefon, | ´az szól? | c.
A
telefon,
az
szól?
‘Is it the telephone that is ringing?’ | A 6telefon, | 6az szól? | When we have said the first MB, that is the subclause in (17) or the non-clausal constituent in (18), we “have said it all”. The second MB brings no complementation and the two MBs in both sentences function as EBs. 5.4.4 5.4.4.1
Complementary blocks (CBs) Syntactic types of complementary blocks
Complementary blocks (CBs) are sister MBs between which there is a complementary relationship. In the present context this relationship
The Melodic Segmentation of Hungarian Utterances
103
means that the non-final CBs in a matrix block are incomplete or “open” without the final or “closing” CB. The open and closing CBs are major-stressed sentence constituents other than ST1. (ST1, as we have seen, constitutes an EB-pair.) The HRS is coextensive with a closing CB, which is divided into its constituent CBs. The rightmost character contour of a non-final CB differs or may differ in certain predictable ways from the rightmost character contour of its final CB sister. This means that the rightmost character contour in a non-final CB is a dependent contour, while the rightmost character contour in the final CB is the dominant contour. Consequently, the non-final CBs are dependent blocks and the final CB is the dominant block. The melody of the non-final CBs depends on the following factors: (a) the dominant contour in the matrix block; (b) the number of major stresses in the non-final CB; and (c) the syntactic type (and pragmatic label if any) of the CB-series (pair). The list below enumerates by name only those syntactic types of CBs that play some special role in melodic segmentation: ST2–ST8. The letters nf and f stand for non-final (open) and final (closing) CBs, respectively. The open CBs in the examples will be italicized.
ST2:
nf = non-final unit of alternative question (i.e. non-final unit of interrogative disjunctive coordination), f = its rightmost sister. e.g. Szombaton, vagy vasárnap? ‘On Saturday or Sunday?’
ST3:
nf = non-final unit of coordination, f = its rightmost sister. Condition: The type of coordination here is different from ST2 and also from ST8 (see below). e.g. Megebédeltek és hazamentek. ‘They had lunch and went home.’
ST4:
nf = clausal constituent (nominal or adverbial subclause) in T or Pv position, f = its rightmost sister. e.g. subclause in T position: T[Amikor megjöttem], még jó volt a lift. ‘When I came, the lift was still working.’; in Pv position: Elküldtem Pv[amit kértél] Debrecenbe. ‘I sent what you had asked for to Debrecen.’
104
ST5:
Intonation and Stress: Evidence from Hungarian
nf = that part of a constituent which is adjacently postmodified by a restrictive relative clause (RRC), f
= the restrictive relative clause.
e.g. akkor, RRC[amikor eloször találkoztunk] ‘when we first met’, literally: ‘then when we first met’; abban a levélben, RRC[amelyiket
én kaptam] ‘in the letter that I received’,
literally: ‘in that letter which I received.’
10
The following three MBs can be realized not only as CBs but also as EBs. We will return to them in section 5.4.5.
ST6:
nf = that part of a constituent which is adjacently postmodified by a non-restrictive relative clause (NRRC) or appositive (App), f
= the non-restrictive relative clause or appositive.
e.g. a levél, NRRC[amit kaptam] ‘the letter, which I received’, a szomszédaiknak, App[Gulyáséknak] ‘to their neighbours, the Gulyáses’. ST7:
11
nf = part of the comment which contains an antecedent in F position which is non-adjacently postmodified by a non-restrictive relative clause or appositive, f = the non-restrictive relative clause or appositive. e.g. F[Az öccsét] hívták be, NRRC[aki diák] ‘They called in his younger brother, who is a student’, where the antecedent is az öccsét ‘his brother-ACC’; F[A szomszédaiknak] adtuk át, App[Gulyáséknak]
‘We delivered it to their neighbours, the
Gulyáses’, where the antecedent is a szomszédaiknak ‘to their 12
neighbours’. ST8:
nf = initial unit of the adversative coordination of the type “x (is), nem (csak) y” = ´x (too), not (only) y´, f
= its rightmost sister.
e.g. Péter, nem Jóska ‘Peter, not Joe’; Péter is, nemcsak Jóska ‘Peter too, not just Joe’.
The Melodic Segmentation of Hungarian Utterances
5.4.4.2
105
Dependent contours before Type α dominant contours
If the rightmost character contour of an HRS is front-falling (i.e. full fall, half fall or fall–rise), we shall call it a Type α dominant contour and symbolize it as αs.
(19) `s |
[s |
7s |
α
s|
Condition: `s | , [s | or 7s | is the rightmost contour of HRS. In CBs, the dependent contour before a Type α dominant contour can be a half fall, a fall–rise or a sustained contour (i.e. rise, high monotone or descent). These possibilities will be accounted for by four melody rules. In the example sentences, contours of Type α will be represented by the full fall, and the sustained contours by the rise. A case where the open CB must have a sustained contour before a Type α dominant contour is where it is the non-final constituent of ST2, that is an alternative question:
(20) a. Péntek vagy szombaton? délután, Friday afternoon or Saturday-on ‘On Friday afternoon, or on Saturday?’ | [Péntek ´délután, | vagy `szombaton? nf. of ST2 b. Pénteken, vagy szombaton? Friday-on or Sat urday-on ‘On Friday, or Saturday?’ | ´Pénteken, | vagy `szombaton? | This is expressed by the CB Melody Rule (i):
(21) CB Melody Rule (i) * * λ α s = obl : s | /([nf (s) (s)0 ])1 [f (Y) s | ] λ where s = sustained character contour (i.e. rise, high monotone, or descent), Y = material of the closing CB up to the dominant (Type α) contour. Condition: nf and f belong to ST2.
106
Intonation and Stress: Evidence from Hungarian
If an open CB does not belong to ST2, and contains more than one majorstressed syllable, its last major-stressed syllable may initiate either a sustained character contour or a half fall before a Type α dominant contour. This is a very general rule, which says that any non-final syntactic constituent that is not an ST1 or ST2 constituent, and which has more than one major stress, may have a sustained final character contour, if the final syntactic constituent MB has a Type α character contour. The rule will be called CB Melody Rule (ii).
(22) CB Melody Rule (ii) * * s = opt : λs | / ([nf (s) (s)1 ])1 [f (Y) αs | ] λ where s = sustained character contour (i.e. rise, high monotone, or descent), Y = material of the closing CB up to the dominant contour (Type α) contour. Condition: nf and f belong to any ST other than ST1 or ST2.
Here are three examples at random:
(23) Amikor megjöttem a moziból, jó volt a lift. when came-1sg the cinema-from good was-3sg the lift ‘When I came home from the cinema, the lift was working.’ | Amikor [megjöttem a ´moziból, | `jó volt a lift. | | Amikor [megjöttem a [moziból, `jó volt a lift. | subclause in T position (24)
Elvittük a
kisebbik lányunkat
szombaton
az állatkertbe.
took-1pl the smaller daughter-our-ACC Saturday-on the zoo-to. ‘We took our younger daughter to the zoo on Saturday.’ | [Elvittük a [kisebbik ´lányunkat | [szombaton az `állatkertbe. | | [Elvittük a [kisebbik [lányunkat [szombaton az `állatkertbe. | nonclausal constituent in Pv position
The Melodic Segmentation of Hungarian Utterances
107
(25) A tanszéki könyvtáros fizetése the departmental librarian salary-his ‘The departmental librarian's salary.’ | A [tanszéki ´könyvtáros | `fizetése. | | A [tanszéki [könyvtáros `fizetése. | modifier within a constituent If an open CB belongs to ST3, ST4, ST5 or ST6, it may have a sustained character contour before a Type α dominant contour, even when it contains only one major-stressed syllable.
(26) Megebédeltek, és hazamentek. had-lunch-3pl and home-went-3pl ‘They had lunch and went home.’ | ´Megebédeltek, | és `hazamentek. | | [Megebédeltek, és `hazamentek. | nf. of ST3 (27) Amikor megjöttem, jó volt a lift. when came-1sg good was-3sg the lift ‘When I came, the lift was working.’ | Amikor ´megjöttem, | `jó volt a lift. | | Amikor [megjöttem, `jó volt a lift. | nf. of ST4, subclause in T position (28) Elküldtem amit kértél Debrecenbe. sent-1sg what asked-2sg Debrecen-to ‘I sent what you had asked for to | [Elküldtem amit ´kértél | `Debrecenbe. | | [Elküldtem amit [kértél `Debrecenbe. | nf. of ST4, subclause in Pv position
Debrecen.’
108
Intonation and Stress: Evidence from Hungarian
(29) Abban, amit that=DEM-in which-ACC ‘In what I received.’ | ´Abban, | amit `én kaptam. | | [Abban, amit `én kaptam. |
én I
kaptam. received-1sg
nf. of ST5
(30) Behívták az öccsét, aki diák. in-called-3pl the younger-brother-his-ACC who student ‘They called in his younger brother, who is a student.’ | [Behívták az ´öccsét, | aki `diák. | | Behívták az [öccsét, aki `diák. | nf. of ST6, f: NRRC
(31) Átadták Gulyáséknak. a szomszédaiknak, over-gave-3pl the neighbours-their-to Gulyáses-to ‘They handed it over to their neighbours, the Gulyáses.’ | [Átadták a ´szomszédaiknak, | `Gulyáséknak. | | [Átadták a [szomszédaiknak, `Gulyáséknak. | nf. of ST6, f: appositive
In certain cases an open CB with a single major stress can have a sustained character contour even when it does not belong to ST3, ST4, ST5 or ST6. This can happen if the CB-series (pair) satisfies certain pragmatic criteria, that is if it is pragmatically marked. A CB-pair receives the pragmatic label “A” (a) if the open CB carries given information and the closing CB new information; or (b) if both the open CB and the closing CB are paralleled or contrasted with their respective counterparts in an explicit or implicit context. For instance:
The Melodic Segmentation of Hungarian Utterances
(32) [Context: Miért feküdt le?] ‘Why did he go to bed?’ mert fáradt volt. Azért feküdt le, that=DEM-for lay-3sg down because tired was-3sg ‘He went to bed because he was tired.’ | ´Azért feküdt le, | mert `fáradt volt. | [Azért feküdt le, mert `fáradt volt. |13 (33) [Context: Elmentek Keszthelyre.] ‘They’ve left for Keszthely.’ voltam. Keszthelyen én is sokat Keszthely-on I too much-ACC was-1sg ‘In Keszthely, I’ve also spent a lot of time.’ | ´Keszthelyen | `én is sokat voltam. | | [Keszthelyen `én is sokat voltam. | (34) [Context: János bácsi újságot olvas.] ‘Uncle John is reading a newspaper.’ tévét néznek. A gyerekek meg the children on-the-other-hand TV-ACC watch-3pl ‘And the children are watching TV.’ | A ´gyerekek meg | `tévét néznek. | | A [gyerekek meg `tévét néznek. | (35) [Context: Mari Rómába ment.] ‘Mary went to Rome.’ És hova Angéla néni? ment and where went-3sg Angela aunt ‘And where did Aunt Angela go?’ | És ´hova ment | `Angéla néni? | 14 | És [hova ment `Angéla néni? | (36) [Context: Lehet, hogy rossz orgonista, de ...] ‘He may be a poor organist, but ... kiváló zongorista. excellent pianist ... he is an excellent pianist.’ ... ´kiváló | `zongorista. | ... [kiváló `zongorista. |
109
110
Intonation and Stress: Evidence from Hungarian
(37) [Context: Nem Jánost küldik, aki tanár, hanem ...] ‘They’re sending not John, who is a teacher, but ... aki diák. Pétert küldik, Peter-ACC send-3pl who student ... they’re sending Peter, who is a student.’ ... ´Pétert küldik, | aki `diák. | ... [Pétert küldik, aki `diák. |15 The possibilities of an open CB with a single major-stressed syllable having a sustained contour before a Type α dominant contour are summed up in (38).
(38) CB Melody Rule (iii) * * s =opt: λs | /([ nf (s) (s)0 ])1 [ f (Y) αs |] λ where s = sustained character contour (i.e. rise, high monotone, or descent), Y = material of the closing CB up to the dominant (Type α) contour. Condition: nf belongs to ST3, ST4, ST5 or ST6; or the CB-series has the pragmatic label “A”. In some cases there is an adversative or concessive–adversative relationship between the non-final and the final CB. In these cases we shall say that the CB-pair receives the pragmatic label “B”. When the CB-pair has the pragmatic label “B”, and the open CB has a single major-stressed syllable, this can initiate a fall–rise (in addition to the half fall) before a Type α dominant contour.
(39) Odamentem, de zárva volt az ajtó. there-went-1sg but closed was-3sg the door ‘I went there, but the door was locked.’ | 7Odamentem, | de [zárva volt az `ajtó. | | [Odamentem, de [zárva volt az `ajtó. | (40) Bár altatót vett be, nem aludt. though sleeping-pill-ACC took-3sg in not slept-3sg ‘Although he took a sleeping pill, he couldn't sleep.’ | Bár 7altatót vett be, | 1nem aludt. | | Bár [altatót vett be, 1nem aludt. |
The Melodic Segmentation of Hungarian Utterances
111
(41) Nem aludt, bár altatót vett be. ‘He couldn’t sleep, though he took a sleeping pill.’ | 7Nem aludt, | bár `altatót vett be. | | [Nem aludt, bár `altatót vett be. | (42) Péterrel nem Jóskával. beszélt, Peter-with talked-3sg not Joe-with ‘He talked to Peter, not to Joe.’ | 7Péterrel beszélt, | `nem Jóskával. | 16 | [Péterrel beszélt, `nem Jóskával. | The rule for this possibility is given in (43).
(43) CB Melody Rule (iv) * α s =opt:7s | /[nf (s) ] [f (Y) s | ] where Y = material of the closing CB up to the dominant (Type α) contour. Condition: the CB-pair has the pragmatic label “B”. 5.4.4.3
Dependent contours before dominant contours other than Type α
If the dominant contour is the stylized fall (see section 3.2.7), all majorstressed syllables in the open CB must also have a stylized fall, independently of the syntactic type or pragmatic label of the CB:
(44) Indulunk Keszthelyre! start-3pl Keszthely-to ‘We are leaving for Keszthely.’ | 4-Indulunk | 4-Keszthelyre! | The rule for this is given in (45).
(45) CB Melody Rule (v) * s =obl: 4-s | /[nf (s) ]([nf (s)4-s | ])0 [f(Y)4-s | ] where Y = material of the closing CB up to the final stylized fall. Finally, let us see CBs whose dominant contour is neither a Type α contour, nor a stylized fall. The dominant contour in these cases will be
112
Intonation and Stress: Evidence from Hungarian
called a Type β dominant contour and we shall symbolize it as βs. Type β contours are the sustained contours, the end-falling contours, and those front-falling contours which are not the rightmost character contours within HRSs:
(46) ´s | , 4s | , 5s | ; 6s | , ¬s | , )s | ; `s | , [s( | ), 7s | β
s
where `s | , [s( | ) , or 7s | is not the rightmost contour of HRS.
When a CB pair has a Type β dominant contour, the dependent contour in the open CB can only be a half falling contour. This does not require a separate rule since all the major-stressed syllables that have been left unaffected by our melody rules will automatically receive a half fall, that being the default melody of pitch accents in Hungarian.
(47) Ha a katonák orzik, ... bejáratot if the entrance-ACC soldiers guard-3pl ‘If soldiers are guarding the entrance, ...’ * | Ha a ´bejáratot | ´katonák orzik, ...| | Ha a [ bejáratot ´katonák orzik, ...| (48) A katonák orzik? bejáratot the entrance-ACC soldiers guard-3pl ‘Are soldiers guarding the entrance?’ * | A ´bejáratot | ˆkatonák orzik? | | A [bejáratot ˆkatonák orzik? | 5.4.5 Melodic blocks that can be used as both complementary and equivalent blocks As was pointed out in section 5.4.4.1, pairs of tonosyntactic blocks that belong to ST6, ST7 and ST8 can be used in two ways: either as complementary blocks or as equivalent blocks, depending on the speaker’s choice. If we wish to convey the idea that the first block requires a continuation, we establish a complementary relationship between the blocks and treat them as CBs. However, if we want to show that we regard the two blocks as pragmatically equivalent, we establish an equivalence relationship between them and treat them as EBs.
The Melodic Segmentation of Hungarian Utterances
113
(49) ST6, where the closing block is a non-restrictive relative clause: a nagyapja, aki fogorvos the grandfather-his who dentist ‘his grandfather, who is a dentist’ a. CB: EB:
...a ´nagyapja, | aki `fogorvos | ...a [nagyapja, aki `fogorvos | ...a `nagyapja, | aki `fogorvos |
b. CB: EB:
...a [nagyapja, aki ´fogorvos, | ... ...a ´nagyapja, | aki ´fogorvos,| ...
c. CB: EB:
...a [ nagyapja, aki 6fogorvos? | ...a 6nagyapja, | aki 6fogorvos? |
(50) ST6, where the closing block is an appositive: a szomszédaiknak, Gulyáséknak the neighbours- their-to Gulyáses-to ‘to their neighbours, the Gulyáses’ a. CB: EB:
...a ´szomszédaiknak, | `Gulyáséknak | ...a [szomszédaiknak, `Gulyáséknak | ...a `szomszédaiknak, | `Gulyáséknak |
b. CB: EB:
...a [szomszédaiknak, ´Gulyáséknak, | ... ...a ´szomszédaiknak, | ´Gulyáséknak, |...
c. CB: EB:
...a [szomszédaiknak, ˆGulyáséknak? | ...a ˆszomszédaiknak, | ˆGulyáséknak? |
(51) ST7, where the closing block is a non-adjacent non-restrictive relative clause: a nagyapja telefonált, aki fogorvos the grandfather-his telephoned-3sg who dentist
114
Intonation and Stress: Evidence from Hungarian
a. CB: | A [nagyapja telefonált, aki `fogorvos | EB: | A `nagyapja telefonált, | aki `fogorvos | ‘It is his grandfather who phoned, who is a dentist’ b. CB: | Ha a [nagyapja telefonált, aki ´fogorvos, | ... EB: | Ha a ´nagyapja telefonált,| aki ´fogorvos, | ... ‘If it is his grandfather who phoned, who is a dentist, ...’ c. CB: | A [nagyapja telefonált, aki ˆfogorvos? | EB: | A ˆnagyapja telefonált, | aki ˆfogorvos? | ‘Is it his grandfather who phoned, who is a dentist?’
(52) ST7,where the closing block is a non-adjacent appositive: a szomszédaiknak adták, Gulyáséknak the neighbours-their-to gave-3pl Gulyáses-to a. CB: | A [szomszédaiknak adták, `Gulyáséknak | EB: | A `szomszédaiknak adták, | `Gulyáséknak | ‘It is to their neighbours that they gave it, the Gulyáses.’ b. CB: | Ha a [szomszédaiknak adták, ´Gulyáséknak, | ... EB: | Ha a ´szomszédaiknak adták, | ´Gulyáséknak, | ... ‘If it is to their neighbours that they gave it, the Gulyáses, ...’ c. CB: | A [szomszédaiknak adták, ˆGulyáséknak? | EB: | A ˆszomszédaiknak adták, | ˆGulyáséknak? | ‘Is it [to] their neighbours [that] they gave it, the Gulyáses?’
(53) ST8, where the closing block is the final unit of the adversative coordination of the type of “x (is), nem(csak) y”: (Ha) (If)
Péterrel beszélt, nem Jóskával Peter-with talked-3sg not Joe-with
a. CB: |7Péterrel [beszélt], | `nem Jóskával.| |[Péterrel [beszélt], `nem Jóskával.| EB: |`Péterrel [beszélt],| `nem Jóskával.| ‘[He talked] to Peter, not to Joe.’
The Melodic Segmentation of Hungarian Utterances
b. CB: |Ha [Péterrel EB: |Ha ´Péterrel ‘If [he talked]
115
[beszélt], ´nem Jóskával, | ... [beszélt],| ´nem Jóskával, | ... to Peter, not to Joe, ...’
c. CB: |[Péterrel [beszélt], ˆnem Jóskával? | EB: |ˆPéterrel [beszélt], | ˆnem Jóskával? | ‘[Did he talk] to Peter, not to Joe?’ Rule (54) is responsible for the intonation of the above examples when they are interpreted as EBs.
(54) EB Melody Rule (ii) * * s =opt ⇒ ωs | /[nf (s) (s)0 ] [f (Y) ωs | ] ω where s = any character contour, equivalent in nf and f. Condition: nf and f belong to ST6, ST7 or ST8. 5.4.6 Summary of the melody rules in the highest-ranking sentence Here is a list of all the melody rules we have established:
(55) Melody Rules a. EB Melody Rule i/ii: * * s → ωs | /[nf (s) (s)0 ] [f (Y) ωs | ] This is obligatory if nf and f belong to ST1, and optional if they belong to ST6, ST7 or ST8. b. CB Melody Rule i/iii: * * s → λs | /([nf (s) (s)0 ])1 [f (Y) αs | ] This is obligatory if nf and f belong to ST2, and optional if they belong to ST3, ST4, ST5, ST6 or have the pragmatic label “A”. c. CB Melody Rule ii: * * s =opt⇒ λs | /([nf (s) (s)1 ])1 [f (Y) αs | ] Here nf and f belong to any ST other than ST1 or ST2.
116
Intonation and Stress: Evidence from Hungarian
d. CB Melody Rule iv: * s =opt⇒ 7s | /[nf (s) ] [f (Y) αs | ] Here nf and f have the pragmatic label “B”.
e. CB Melody Rule v: * s =obl 4-s | /[nf (s)
]([nf (s)4-s | ])0 [f (Y)4-s | ]
Here are the steps in which we can establish the non-final melodies within any highest-ranking sentence (HRS). (i) Check whether or not the constituent MBs of the matrix block are (treatable as) equivalent blocks. a. If they are equivalent blocks because they belong to ST1, or you regard them as equivalent blocks because they belong to ST6, ST7 or ST8, and so they entitle you to have this option, then provide the last major-stressed syllable in the non-final MB with the same contour as the final character contour of the matrix block, see rules (16) and (54). b. If they are not (or in the case of ST6, ST7 or ST8 you do not choose to treat them as) equivalent blocks, then they are complementary blocks and you should consider the final character contour of the matrix block. (ii) Check whether or not the final character contour of the matrix block is a Type α contour (i.e. a full fall, half fall or fall–rise which is simultaneously the rightmost contour of the HRS). a. If it is, you can do the following. If the constituent MBs of the matrix block belong to ST2, then provide the last majorstressed syllable in the non-final constituent MB with a sustained contour, see rule (21). If a non-final constituent MB has more than one major stress, then you may provide the last major-stressed syllable in this constituent with a sustained contour, see rule (22). If the constituent MBs belong to ST3, ST4, ST5, or if they have the pragmatic label ″A″, or if they belong to ST6, ST7 or ST8 and you want to treat them as complementary blocks, then you may provide the last majorstressed syllable in the non-final constituent MB with a sustained contour, see rule (38). If the constituent MBs have the pragmatic label ″B″, then you may provide the last major-
The Melodic Segmentation of Hungarian Utterances
117
stressed syllable in the non-final constituent MB with a fall– rise; see rule (43). b. If the final character contour of the matrix block is a stylized fall, then provide all pitch-accented syllable sequences in the HRS with stylized falls, see rule (45). In all other cases (i.e. if the final character contour of the matrix block is a Type β contour), the major-stressed syllables in the matrix block will receive a half fall (by default). 5.4.7
Contour replacements
When the melodies have been assigned by the melody rules, some melodic replacements may take place, for attitudinal purposes. The first of these is the replacement of a half fall with a fall–rise, before a dominant full fall. The resulting fall–rise does not imply a conflict, it simply serves as a special emphasis marker on non-final constituents and signals for the listener that the speaker is going to continue. This replacement can happen in emphatic rhetorical style.
(56) Replacement of half fall by fall–rise ])1 [f (Y) `s | ] [s=opt: 7s | / ([nf (s)([s)0 where Y = material of the closing CB up to the dominant full fall. Condition: emphatic rhetorical style. NB: When the fall–rise replaces a scalar half fall, the replacement automatically splits the IP into two. Example (57), taken from a priest’s sermon, shows this kind of replacement on the word Szentírás ‘Holy Scripture’.
(57) Maga a Szentírás figyelmeztet minket erre. itself the Holy-Scripture warns us this-on ‘The Holy Scripture itself warns us of this.’ 18 | [Maga a 7Szentírás | [figyelmeztet 4minket | `erre. | The other kind of replacement affects the series of half fall, half fall, full fall, and changes the medial half fall into a high monotone, with a change in the attitude conveyed. Since the high monotone is a terminal IP, this operation automatically splits the original IP into two. For instance, (58)n, which is attitudinally unmarked, can be replaced in this way by the attitudinally marked (58)a:
118
Intonation and Stress: Evidence from Hungarian
(58) Mert nem volt elegendo bizonyítékom. because not was-3sg enough proof-my ‘Because I did not have sufficient evidence.’ The normal version: n. |Mert [nem volt nem Mert
[elegendo
`bizonyítékom.|
e volt
legendo
bi zonyíté
kom.
Replacement by the high monotone: a. |Mert [nem volt 4elegendo | `bizonyítékom.| nem Mert
elegendo bi volt
zonyíté kom.
Using Bolinger’s (1986, 1989) terms, this kind of contour replacement changes a sequence of A + A + A profiles into a sequence of A + B + A profiles over the same string of syllables. The sequence of half fall, high monotone, full fall at the end of an utterance so obtained signals that the speaker considers the contents of the sentence self-evident (i.e. obvious, natural, easy to guess or easy to understand), usually with an implication of the speaker’s intellectual superiority or official authority and it sounds patronising and categoric at the same time but is not unfriendly. Consequently it is especially common in lecturing and explanations given to groups of children by teachers, but it also occurs in individual self-defence and in utterances repeated because they were misheard. I shall refer to this complex attitude as “schoolmasterly”. As a first approximation, this kind of contour replacement can be defined as follows.
(59) Replacement of half fall by high monotone (Preliminary) [s =opt: 4s | /[s `s | Condition: “schoolmasterly” attitude.
The Melodic Segmentation of Hungarian Utterances
119
Further examples show that our first definition of this kind of contour replacement is not general enough because it does not cover cases like (60)b, although that, too, is a possible contour-replaced version.
(60) Ráborította a hamutartót a frissen festett asztalra. on-spilt-3sg the ashtray-ACC the freshly painted table-onto ‘He spilt the ashtray on the newly painted table.’ The normal version: n. |[Ráborította a [hamutartót a [frissen festett `asztalra.| Rá ha fris borította a mutartót a asz sen festett talra.
Replacement by the high monotone: a. |[Ráborította a [hamutartót a ¯frissen festett | `asztalra.| Rá ha frissen festett asz borította a mutartót a talra. b. |[Ráborította a ¯hamutartót a | [frissen festett `asztalra.| Rá hamutartót a fris borította a aasz sen festett talra.
We can account for cases like (60)b if we allow replacement by the high monotone to happen before a final CB of any depth, that is not only before asztalra, but also before a frissen festett asztalra, see (60)n′:
120
Intonation and Stress: Evidence from Hungarian
(60)n’ Ráborította a hamutartót a frissen festett asztalra.
CBnf
CBnf
CBnf
CBf
CBf
CBf = Sentence This discovery calls for a more general definition of replacement by the high monotone, which is given in (61).
(61) Replacement of half fall by high monotone (Extended) [f(Y) `s|] [s =opt: 4s| /[s where Y = material of the closing CB up to the falling contour. Condition: “schoolmasterly” attitude. The extended rule, applied iteratively, permits replacement by the high monotone not only on the syllable fris-, as in (60)a, but also on the syllable ha-, as in (60)b. Moreover, it permits replacement in both places simultaneously, as in (60)c:
(60)c. | [Ráborította a 4hamutartót a | 4frissen festett| `asztalra.| Rá hamutartót a frissen festett borította a asz talra. However, it has to be noted that the output of the contour replacement rule, (61), may coincide with the output of CB Melody Rule ii, (22), introduced in section 5.4.4.2 above. The latter rule says that when a non-final CB contains more than one major stress, its last major-stressed syllable may have a sustained contour (including the high monotone) if the closing CB has a Type α dominant contour (including the full fall). Consequently, in a case like (62) it is impossible to tell (from the intonation alone) whether the high monotone in the middle is due to (61), with the “schoolmasterly” attitude that it conveys, or whether it is due to Rule (22), without such an attitude.
The Melodic Segmentation of Hungarian Utterances
121
(62) Magyarország keleti részén esik az eso. Hungary eastern part-its-on falls the rain ‘In the eastern part of Hungary it is raining.’ | [Magyarország 4keleti részén | [esik az `eso. | It seems that rules (22) and (61) conspire to achieve the prosodic solution of cases like (62). 5.4.8
Pauses in the highest-ranking sentence
Syntax-based pauses are also assigned in the stage of the melodic segmentation of the highest-ranking sentence. At a normal rate of speech, a pause is most likely at the end of the HRS. Within the HRS it is quite likely after a preposed constituent (ST1), after a unit of coordination (ST2, ST3, ST8), after a nominal or adverbial subclause (ST4), and before a non-restrictive relative subclause or apposition (ST6, ST7). It often occurs after a non-clausal T constituent, and before a restrictive relative clause that postmodifies its antecedent non-adjacently; less often before a restrictive relative clause that postmodifies its antecedent adjacently (ST5). It is not likely between modifier and head within major syntactic constituents unless the modifier is a subclause or appositive, when a pause before the modifier is possible. All character contours apart from the half fall end in an IP boundary, no matter whether or not they end in a pause. Consequently, it is only the half-falling character contour whose status is changed by the addition of a final pause. A pause at the end of a half fall turns the scalar half fall into a terminal one, that is it creates a new IP boundary at the end of the half fall (see section 3.6):
(63) Pétert küldik, aki diák. Peter-ACC send-3pl who student ‘They are sending Peter, who is a student.’ a. | [Pétert küldik, aki `diák.| b. | [Pétert küldik, | - aki `diák.| Pauses in other positions do not affect the IP boundaries already established:
(64) Pénteken, vagy szombaton? Friday-on or Saturday-on ‘On Friday, or Saturday?’ a. |´Pénteken, | vagy `szombaton? | b. |´Pénteken, | -vagy `szombaton? | -
122
5.5
Intonation and Stress: Evidence from Hungarian
Deviations
The rules of melodic segmentation shown in sections 5.3 and 5.4 manifest themselves directly in sentences that are planned before they are uttered. In the spontaneous speech of actual performance, however, we normally form our sentences on line, that is while we are uttering them, and so they are often deviant syntactically and/or intonationally. When we change our sentences in the course of saying them, there may be a conflict between the melodies we assign before and after these changes, and the contour sequences actually produced may become “deviant” (with respect to the rules we have established). For instance, while (65)a is non-deviant melodically because it obeys all the relevant melody rules, (65)b is deviant: it contains a “wrong” contour sequence, violating some of the melody rules.
(65) Elviszem a legjobb gyerekeket az állatkertbe take-1sg the best children-ACC the zoo-to és veszek nekik fagylaltot is. and buy-1sg them ice-cream-ACC too ‘I’ll take the best children to the zoo, and buy them some ice-cream, too.’ a. | Elviszem a [legjobb [gyerekeket az ´állatkertbe, | és [veszek nekik `fagylaltot is.| b. *? | [Elviszem a [legjobb ´gyerekeket | az ´állatkertbe, | és [veszek nekik `fagylaltot is.| Nevertheless, it is possible to find reasons why (65)b can occur in performance. It may happen, for example, that the speaker does not at first think of the continuation és veszek nekik fagylaltot is ‘and buy them some ice-cream, too’, and wishes to drop his/her voice on the word állatkertbe. This justifies the rising contour on gyerekeket, because a legjobb gyerekeket is a non-clausal constituent in Pv position, containing more than one major-stressed syllable, and so rule (22) can apply. The utterance is made intonationally deviant by the subsequent addition of the final coordinated clause, which was not originally planned, and because of which the contour on állatkertbe also becomes rising. Nevertheless, as syntactically deviant utterances do not invalidate syntactic rules, intonationally deviant utterances do not invalidate melody rules. Changing the sentence in the course of saying it by inserting originally unplanned additions may result in the previous melody assignment becoming superfluous or odd. But normally we do not
The Melodic Segmentation of Hungarian Utterances
123
bother to correct these intonational “mistakes” because they have a role to play: they indicate hesitation by signalling to the hearer that the speaker has changed the utterance in the middle of producing it. Another possible reason for the occurrence of (65)b may be that the speaker first wants to say: Elviszem a legjobb gyerekeket és veszek nekik fagylaltot is ‘I’ll take the best children out and buy them some ice-cream, too’. The end of the first coordinated clause in this version (gyerekeket) receives a rising contour because elviszem a legjobb gyerekeket is a coordinated constituent belonging to ST3. Then, before going on with the second clause, the speaker adds az állatkertbe to the first clause, as an afterthought, also with a rising contour. Here the afterthought az állatkertbe appears in the middle of a structural-functional sentence. But afterthoughts are more frequent after structural-functional sentences, either within the same utterance, as in (66)a, or in separate utterances, as in (66)b. The phrase most se ‘not even now’ is an afterthought in both examples:
(66) Kispárna nem kell, most small-cushion not necessary now ‘Don’t you want a pillow, even now?’ a. | [Kispárna ˆnem kell, | ˆmost se? | b. | [Kispárna ˆnem kell? | - ˆMost se? |
se? neither
And the subsequently added parts can not be only afterthoughts, but modifications, corrections, too, as in (67):
(67) És ha nem akarják? Vagy nem lehetséges? and if not want-3pl or not possible ‘And [what] if they don’t want it? Or if it is not possible?’ | És ha ´nem akarják? | - Vagy ´nem lehetséges? |
Whenever we subsequently add something to an already realized utterance or utterance-part, we can signal the relatedness between the added part and the previous utterance(-part) by assigning the same character contour to the last major-stressed syllable in both. That is to say, we carry out left-to-right contour concord between the utterance(-part) and the addition. However, this is not a melody rule, it is only a strategy belonging to performance.
124
5.6
Intonation and Stress: Evidence from Hungarian
Summary
This chapter has examined how Hungarian utterances divide into intonational phrases. This division is determined partly by the syntactic, partly by the prosodic structure of the utterance. We first discussed the intonation of major tonosyntactic blocks, which are melodically autonomous blocks within utterances. Then we examined the intonation of minor tonosyntactic blocks (i.e. major-stressed blocks into which the major tonosyntactic blocks can be decomposed) and found that their melodies show right-to-left dependencies which can be expressed in melody rules. Since these rules are related to the surface syntax of utterances, they manifest themselves particularly well in pre-planned speech or careful reading, but may be violated in unplanned spontaneous speech, where melodically deviant utterances may occur. But even the deviations can only be explained by referring to the melody rules. And as the (often) deviant syntax of utterances does not invalidate syntactic rules, deviant intonation does not invalidate the rules of intonation either.
Part II Stress
This page intentionally left blank
6 Stress in Hungarian Words, Phrases and Sentences
6.1
Introduction
Since intonation and stress are closely interrelated, it is necessary for us to examine the stress system of Hungarian in some detail. In section 6.2 we briefly recapitulate the phonological stress degrees of Hungarian we introduced in Chapter 3 and give our reasons for distinguishing these three degrees. In section 6.3 we shall examine stress in Hungarian words, including ordinary compounds. (The stressing of phrasal compounds will be discussed separately in Chapter 7.) Finally, in section 6.4. we shall consider the stressing of phrases and sentences, that is the stage of stress fixing in the syntax–phonology mapping (see section 1.2.2).
6.2
The phonological stress degrees of Hungarian
In section 3.2.1 we sketched the Hungarian stress system as having three phonological degrees: major, minor and zero. Major-stressed syllables are marked by the simultaneous presence of two kinds of prominence: (a) melodic – they initiate one of the character contours of Hungarian; and (b) non-melodic – they have extra intensity as opposed to the intensity of unstressed syllables. Because of the first kind of prominence, major-stressed syllables are said to be pitch-accented or accented. The intonation contours that begin on such syllables lend special prominence to these syllables. Minor-stressed syllables have only non-melodic prominence, that is extra intensity, as opposed to the intensity of unstressed syllables, and optionally a slight pause or juncture before them, but they do not initiate character contours (i.e. do not have a pitch accent). 127
128
Intonation and Stress: Evidence from Hungarian
Zero-stressed (or unstressed) syllables have neither a pitch accent, nor extra intensity. While major-stressed syllables are accented, minor- and zero-stressed ones are unaccented. At the same time, both major- and minor-stressed syllables start feet, while zero-stressed syllables do not. The capacity of stressed syllables to start feet is a corollary of their extra intensity. Within words, major stresses are usually referred to as primary, minor stresses as secondary stresses. Vowel quality does not play a role in distinguishing stress degrees in Hungarian. Hungarian vowels typically have a full, non-reduced quality, even in zero-stressed syllables; see, for example, Kerek (1971: 36–7). 1 The lack of systematic qualitative reduction may occasionally cause some difficulty in perceptually distinguishing minor and zero stresses from each other. Therefore other kinds of evidence may also have to be looked for in order to be able to identify minor stresses. 2 Nevertheless, in general we are able to produce and hear clear cases of minor stress and therefore the phonetic reality of Hungarian minor stress has not been brought into question. For the notation of stress we can choose either of two methods. We can either transcribe stress and intonation simultaneously, by means of tonetic stress marks of the kind we introduced in Chapter 3, as in, for example, (1)a; or we can transcribe stress only, by means of the symbols [[] and []] for major and minor stress, respectively, as in (1)b.
(1) Megjött Angéla néni. PERF-came-3sg Angela aunt a. | [Megjött 1Angéla ]néni. | b. [Megjött [Angéla ]néni.
‘Aunt Angela has arrived.’
As we saw in Chapter 3, tonetic stress marks like the ones used in (1)a indicate major stress and character contour simultaneously, for example [[] and [1] indicate the initiation of a half fall and a full fall, respectively, and occur in combined transcriptions of stress and intonation. Such transcriptions are always given between two [|] symbols, that is between IP boundaries. In addition to the tonetic stress marks, such transcriptions may also contain a symbol for minor stress []], a symbol for the appended contour [3] and a pause symbol [-]. However, when we are uninterested in the actual melodies which the major stresses initiate, we can use a simple stress transcription, as in (1)b. The three degrees of stress are prosodemes because they can be in contrast, as shown in (2)a, (2)b and (2)c.
Stress in Hungarian Words, Phrases and Sentences
129
(2) a. [Context: Az idén nincs hó.] ‘There is no snow this year’. [Január [hoz. January brings ‘January will BRING some.’ b. [Context: Február hoz havat.] ‘February will bring some snow.’ [Január ]hoz. ‘JANUARY will bring some.’ c. [Context: Melyik hónaphoz írtad be?] ‘Which month did you write it to?’ [Januárhoz. January-to ‘To January.’
6.3
Hungarian stress is not fully relational
The three degrees of stress I have presented are discrete categories of prominence because the presence or absence of pitch accent (character contour initiation) and the presence or absence of non-melodic prominence (extra intensity) are matters of an either-or nature, and the actual amount of the extra intensity forms no part of the definition. This was summed up in Table 3.1, reproduced here as Table 6.1, for the reader’s convenience. As can be seen from Table 6.1, stress degrees are treated here as partly relational, partly absolute categories. They are relational inasmuch as minor- and major-stressed syllables have more intensity than the unstressed syllables (i.e. both are “stressed”, as opposed to the “unstressed” syllables), but they are absolute inasmuch as major-stressed syllables initiate character contours (and so carry a pitch accent), while the other Table 6.1
The system of Hungarian stress prosodemes
Zero stress Minor stress Major stress
Extra intensity
Initiation of character contour
– + +
– – +
130
Intonation and Stress: Evidence from Hungarian
syllables do not. Consequently, the only intensity difference relevant to the present framework is the one that is detectable between unstressed syllables on the one hand and non-unstressed (i.e. major- or minorstressed) syllables on the other. A major-stressed syllable therefore is not necessarily more intensive than a minor-stressed one (but it is more prominent because of the pitch accent it carries), and two major-stressed syllables may have different degrees of intensity as long as both have a larger degree of intensity than the unstressed syllables do. According to some views (cf. É. Kiss 1987, 1992; Vogel 1987), the lefthand constituent in a binary Hungarian phrase is stronger (i.e. its stressed syllable has more intensity) than the right-hand constituent, except when the left-hand constituent is topic and the right-hand constituent is comment; is this case the topic is weaker. (For the notions of topic and comment, see sections 1.2.3 and 5.2.) According to these views the Hungarian Nuclear Stress Rule works in the opposite direction to its English counterpart. For instance, in (1) above, reproduced here as (3), the syllable An-, which starts the terminal full fall, is said to be weaker than the syllable meg-, which starts the half fall. That is to say, the strongest syllable (“designated terminal element” or DTE) and the syllable starting the terminal part in the Hungarian IP (see section 3.6.1) do not have to coincide. Under this analysis, the syllables meg- and Anare explicitly treated as either a stronger and a weaker kind of major stress, or a major stress and a minor stress, respectively.
(3) | [Megjött 1Angéla ]néni. | In our framework, however, the slightly less physical prominence of the syllable An- does not warrant the view that the syllable An- has only minor stress or a weaker kind of major stress that is worth distinguishing. Since the syllables meg- and An- both satisfy the criteria of majorstressedness (i.e. both of them are pitch-accented), both of them are major-stressed, and any differences in their intensities are regarded as merely phonetic. Moreover, proponents of the Hungarian Nuclear Stress Rule assert that, since one syntactic phrase can be embedded in another, several degrees of major stress intensity will occur. For instance, the phrase in (4) has four major-stressed syllables, and they are said to have three degrees of intensity, with the leftmost one being the strongest, the third one being the second strongest, while the second and fourth ones rank equal as third strongest; cf. É. Kiss (1992: 94).
Stress in Hungarian Words, Phrases and Sentences
131
(4) Esterházy Péter legújabb könyve Esterházy Peter newest book-his ‘Peter Esterházy’s latest book’ a. | [Esterházy [Péter [legújabb 1könyve. | b. [Esterházy [Péter [legújabb [könyve. But again, in our framework, they all satisfy the definition of major stress because all four of them, including the weakest ones, are associated with pitch accents. In the present discussion, the phonetic differentiation of major stresses according to their intensity is irrelevant. As a matter of fact, the plausibility of a syntax-based differentiation of major stresses in Hungarian is somewhat doubtful, anyway. It was rejected by Kálmán and Nádasdy (1994: 405–6) on the grounds that the syntax-based stress computations related to the Nuclear Stress Rule often give incorrect results. Fónagy (1998: 340) expresses a similar view: “[f]or the utterances of the unmarked SUBJECT (Nominal) + PREDICATE (Verbal) type, such as Kati futott (Kati was running), or Péter tudta (Peter knew it), the accent strikes the subject and the predicate with equal force. This makes it impossible to apply the metrical grid analysis proposed by Mark Liberman.” That is to say, Hungarian major stresses are often “roughly equivalent” in strength, even in phonetic terms. Although at the phonetic level major stresses may or may not occur with equal strengths, it is our contention that from a phonological point of view all major stresses should be regarded as equal and their intensity differences (if indeed there are any) should be ignored as phonologically irrelevant. This follows not only from our definition of major and minor stress but it is also supported by the facts of rhythmical variation in Hungarian, described in Chapter 7.
6.4
Representation
The consequence of the preceding discussion is that major stresses need not be phonologically distinguished according to their strengths. This eliminates the necessity of using a metrical grid to represent Hungarian metrical structure. Even if we did use such a grid, we would not distinguish major stresses according to their intensity, and consequently we would ignore the levels above Line 3 in our grids. That is to say, the relevant “window” in each of our grids would include only the bottom three levels: the columns of major-stressed syllables would always contain three grid marks, those of minor-stressed syllables would contain two, and those of unstressed syllables would contain one grid mark. This
Intonation and Stress: Evidence from Hungarian
132
possibility is shown in (5)a. Instead of such grids, however, we shall use a much simpler metrical representation: an unmatched left parenthesis before each (minor- and major-) stressed syllable and an asterisk (accent mark) * over each major-stressed syllable, as in (5)b. As we shall see in the next section, the unmatched left parenthesis always marks the beginning of a foot. The accent mark cannot occur without the left parenthesis, but the left parenthesis can occur without the accent mark.
(5) = (1) [Megjött [Angéla ]néni. ‘Aunt Angela has arrived.’ a. Line 3 x x Line 2 x x x Line 1 x x x xx x x Megjött Angéla néni. b.
6.5
* * (Megjött (Angéla (néni.
Stress in Hungarian words
6.5.1
Stress in simple and derived lexical words
When talking about stress within words, it is customary to refer to the major stress as primary stress or main stress and to the minor stress(es) as secondary stress(es). Hungarian lexical words (i.e. non-function words), whether simple, as (6)a, or derived, as (6)b, have a single primary stress, which falls on the first syllable of the word, and they have no secondary stresses:
(6)
a. [pálinka b. [matematikus
‘brandy’ ‘mathematican’
3
(simple word) (derived word)
In terms of prosodic constituents, I equate each non-compound (i.e. simple or derived) lexical word with an unbounded foot (Ft). This is a left-headed foot, starting on the initial, primary-stressed syllable of the word. 4 In my view, each lexical word is furnished with a primary stress (accent) on its first syllable and, concomitantly, with the left boundary of an unbounded foot before its first syllable. This happens in the lexical stratum of the phonology and is achieved by the Word Stress Rule, (7). The accent is indicated by the asterisk *, and the left foot boundary
Stress in Hungarian Words, Phrases and Sentences
133
by a left parenthesis. Adding the asterisk automatically involves adding the left parenthesis.
(7) Word Stress Rule (WSR) (preliminary) *
σ =obl⇒ (σ / W[
σ0] where W is a lexical word. The rule has been given here in a preliminary form, its scope will be extended in section 6.5.3. As a result of its application, (6)a and (6)b will leave the lexical stratum as (8)a and (8)b:
(8) a.
* (pálinka
b.
* (matematikus
When the word becomes part of a phrase or sentence, that is when it enters the postlexical stratum of phonology, it will bring its primary stress and left foot boundary with it. Since the primary stress is fixed on the first syllable of Hungarian words, its position has a demarcative and not a distinctive function in Hungarian. It cannot distinguish the meaning or category membership of words as it can in English (see, for example, [importN vs. im[portV ). On the other hand, it can distinguish a word from a combination of two words built up of the same segmental phonemes, if the first word in the combination is a zero-stressed clitic: [halát ‘his fish-ACC’ vs. ha [lát ‘if he can see’, or [akar ‘he wants to’ vs. a [kar ‘the arm’. (For clitics, see section 6.5.3.) 6.5.2
Stress in lexical compounds
We have to make a distinction between ordinary and phrasal lexical compounds. In ordinary lexical compounds, the first syllable of the first constituent is primary-stressed and the first syllable of the second constituent is secondary-stressed in Hungarian, see (9)a. By contrast, phrasal compounds have an exceptional, phrase-like stressing: they retain a primary stress on each of their immediate constituents, see (9)b.
(9)
a. [csónak]verseny boat race
‘boat race’
(ordinary compound)
b. [negyven[nyolc 40 eight
‘48’
(phrasal compound)
5
134
Intonation and Stress: Evidence from Hungarian
The secondary stress in ordinary compounds is derived through obligatory reduction from the primary stress, by the application of the Compound Stress Rule, given in (10).
(10) Compound Stress Rule (CSR) * * σ =obl: σ / W[ A[(σ σ0 ] B[(
σ0 ] ]
In a configuration W[AB], where A, B and W are lexical categories, deaccent B, unless W is a phrasal compound. Removing the accent from the first syllable of constituent B does not affect the foot boundary before it, and so the first syllable of B remains stressed, only its stress degree is reduced to secondary. The stressing of phrasal compounds does not need a separate rule. The negative condition in (10) ensures that, by not carrying out the CSR, the phrasal compound will retain all its primary stresses. At this point we suspend further discussion of phrasal compound stressing and postpone it until Chapter 7. In the rest of this section we shall deal solely with the stressing of ordinary compounds, and the word compound will be used to mean ‘ordinary compound’. As for compounds which contain embedded compounds, Kálmán and Nádasdy (1994: 408) imply that it is only the highest (non-embedded) constituents of the compound that receive stress; the more deeply embedded constituents are zero-stressed, see, for example, the two versions of pótkerékcsapágy in (11)a and (11)b.
(11) a.
pót kerék csap spare wheel pin [pót]kerékcsapágy
ágy bed
‘spare wheel-bearing’
pót kerék csap 6 [pótkerék]csapágy
ágy
‘spare-wheel bearing’
b.
Kálmán and Nádasdy (1994: 408)
Stress in Hungarian Words, Phrases and Sentences
135
I suggest two obligatory lexical rules to account for this difference: the Compound Stress Rule, (10), and Lexical Foot Erasing (LFE), (12). Lexical Foot Erasing is ordered before the Compound Stress Rule.
(12) Lexical Foot Erasing (LFE) * ( =obl⇒ 0 / W[ A[( σ σ0 ] B[
σ σ0 ] ]
In a configuration W[ A B ], where A, B and W are lexical categories, remove the foot boundary at the beginning of B, that 7 is destress B. Comment: ordered before the CSR. Since LFE precedes (and feeds) CSR and since its environment is created by an application of the CSR on a previous cycle, it will delete the left parenthesis before all and only the non-immediate constituents of the compound. The derivation of (11)a and (11)b can be given as (13)a and (13)b, respectively.
(13) a. Derivation of [pót]kerékcsapágy * * * * (i) [(csap(ágy] 3 [(csap(ágy] 3 [(csapágy ] CSR LFE * * * * (ii) [(kerék(csapágy] 3 [(kerék(csapágy] 3 [(kerékcsapágy] CSR LFE * * * (iii) [(pót(kerékcsapágy] 3 [(pót(kerékcsapágy] CSR b. Derivation of [pótkerék]csapágy * * * * * * (i) [(pót(kerék] [(csap(ágy] 3 [(pót(kerék] [(csap(ágy] 3 CSR LFE * * [(pótkerék] [(csapágy] * * * (ii) [(pótkerék(csapágy] 3 [(pótkerék(csapágy] CSR The derivations in (13)a and (13)b show that the LFE cannot be applied before the first application of the CSR because its structural description
136
Intonation and Stress: Evidence from Hungarian
is not met there, and neither can it be applied after the final application of the CSR because it is ordered before the CSR. Both the CSR and the LFE apply within the lexical stratum of the phonology, and so each ordinary compound enters the postlexical stratum with one primary stress (accent) and as many unbounded feet as the number of its immediate constituents.8 It is likely that the representation in (11)a, [pót]kerékcsapágy, which is the output of (13)a, will undergo a postlexical change: Clash Deletion. This rule, given in (39) below, deletes the minor stress of a syllable which is immediately after a stressed syllable. If Clash Deletion applies to (11)a, the actual pair pronounced will be (14)a and (14)b, rather than (11)a and (11)b:
(14)
a. [pótkerékcsapágy b. [pótkerék]csapágy
9
‘spare wheel-bearing’ ‘spare-wheel bearing’
Moreover, all the minor stresses (including the secondary stresses within compounds) in Hungarian utterances can be deleted (see Kálmán and Nádasdy 1994: 409, Varga 1998: 230), by the optional application of Ultimate Foot Erasing:
(15) Ultimate Foot Erasing (UFE) (postlexical) σ σ0 ] ... ] ( =opt: 0 / IP[... W[ In an intonational phrase, optionally remove the foot boundary at the beginning of each W, where W is an unaccented lexical word. Domain: intonational phrase. The UFE takes last place in the order of rules within the postlexical stratum. In principle, the deletion it causes may happen to all of the minor stresses or none of them, in one single operation. But it seems more likely that the rule applies to the minor stresses one by one and, since it is optional, it may delete some of the minor stresses and spare others, at the speaker’s discretion.10 For instance, if it is allowed to operate on (14)b, the difference between (14)a and (14)b is completely neutralized and both of them become [pótkerékcsapágy. 6.5.3
Stress in function words
While lexical items conveying new information within Hungarian phrases and sentences are normally major-stressed, function words are not. The latter are divided with respect to their stressing: some of them are typically zero-stressed, some are typically minor-stressed.
Stress in Hungarian Words, Phrases and Sentences
137
The ones that are typically zero-stressed are clitics. Therefore, I claim that clitics leave the lexical stratum with zero stress, but the rest of the function words leave the lexical phonology with major stress, just like lexical words do, and lose this major stress postlexically, through deaccentuation. By clitics I mean monosyllabic function words that are typically zero-stressed. They include articles, all of which are monosyllabic in Hungarian: a(z) ‘the’11, egy ‘a(n)’; monosyllabic conjunctions, for example és ‘and’, de ‘but’, vagy ‘or’, is ‘also’, ha ‘if’; monosyllabic personal and demonstrative pronouns, for example én ‘I’, o ‘(s)he’, ez ‘this’; certain monosyllabic adverbs, for example már ‘already’, még ‘still’, csak ‘only’.12 We assume that all clitics (whether they are enclitics or proclitics) become part of the unbounded foot started by the preceding word. Since these feet are not closed at their right edges, the clitics following them get incorporated into them automatically, and no separate rule is needed to achieve this. If there is no unbounded foot before the clitic, the clitic remains unattached (unfooted). For instance, the initial definite article in (16) is unattached.
[csónak]verseny boat race boat race and * (csónak(verseny és
(16) a the ‘the a
és a [barátom. and the friend-my my friend’ * 13 a (barátom.
Other function words are non-clitic function words. These are polysyllabic (e.g. conjunctions like mindazonáltal ‘nevertheless’; reflexive pronouns like magamat ‘myself-ACC’, maguktól ‘from themselves’; relative pronouns like aki ‘who-NOM’, amelyikrol ‘about which’; polysyllabic inflectional forms of personal pronouns like benneteket ‘you-pl-ACC’, hozzátok ‘to you-pl’, toletek ‘from you-pl’, etc.; or polysyllabic inflectional forms of demonstratives like ebbol ‘from this’, ezekhez ‘to these’, etc.). When such words appear in phrases or sentences, their first syllable is typically minor-stressed, and may even be major-stressed when the utterance is made in a special narrative style (Kálmán and Nádasdy 1994: 462).14 In (17) the non-clitic function word benneteket is minor-stressed.
(17) [Meghívlak benneteket invite-1pl you-pl-ACC ‘I’ll invite you to dinner.’ * *
[vacsorára. dinner-to
138
Intonation and Stress: Evidence from Hungarian
(Meghívlak (benneteket (vacsorára. | [Meghívlak ]benneteket 1vacsorára. | The non-clitic function words are not unstressed.15 One may object that the prominence-increase which we believe to be a minor stress on such words is merely a phonetic fact and that, phonologically, these words are also zero-stressed. But apart from the fact that this view cannot account for the major-stressed cases which occur in the special narrative style mentioned above, there is another reason why it cannot be accepted. The prominence-increase on the function word is a real minor stress because it is capable of receiving Contour Insertion, which we can use as a diagnostic test for minor stress, see Chapter 8, section 8.3. Example (18) shows the contour-inserted version, see (17):
(18)
* * * (Meghívlak (benneteket (vacsorára. | [Meghívlak 4benneteket | 1vacsorára. |
So the minor stress on the first syllable of benneteket in (17) is a genuine minor stress. Since it is not rhythmical (see Chapter 8), it must have some other source. We shall assume that such words are supplied with primary stress and an unbounded foot boundary in the lexical part of the phonology, just like lexical words are. To account for this possibility, rule (7) has to be revised in the following way:
(19) Word Stress Rule (WSR) (Revised) *
σ =obl⇒ (σ / W[
σ 0] where W is a lexical word or a non-clitic function word. When the non-clitic function words are inserted in phrases and sentences, they will undergo a kind of obligatory deaccentuation, that is their major stress will be reduced to minor (see next section).
6.6 6.6.1
Stress in Hungarian phrases and sentences Syntactically motivated deaccentuation
The stressing of phrases and sentences is the result of the first stage in the syntax–phonology mapping: stress fixing (see section 1.2.2). It is this that we shall now examine.
Stress in Hungarian Words, Phrases and Sentences
139
In Hungarian phrases all lexical words that convey new information retain their major stress, and so several, roughly equal, major stresses can occur, as in for example (20).
(20) [Esterházy [Péter [legújabb Esterházy Peter newest ‘Peter Esterházy’s latest book’ * * * (Esterházy (Péter (legújabb
16
[könyve. book-his
* (könyve.
The same applies to non-emphatic Hungarian declarative sentences, but the verb in these loses its major stress when the F position is filled, as in (21). This sentence illustrates all the basic syntactic positions identified in the Hungarian syntactic model (É. Kiss 1981, 1987, 1992) introduced in section 1.2.3. The structure of the sentence is shown in (21)a, its stressing in (21)b and its division into unbounded feet in (21)c.
(21) János mindig kávét iszik a földszinti John always coffee-ACC drinks the groundfloor büfében. snackbar-in ‘John always drinks coffee in the groundfloor snackbar.’ a.
S XP
VP XP
VP V′
XP Jánosi mindigj
V
XP
XP XP XP
kávétk iszik a földszinti büfében Q T
F
V
0i
0j
Pv C
b. [János [mindig [kávét ]iszik a [földszinti [büfében. * * * * * c. (János (mindig (kávét (iszik a (földszinti büfében.
17
0k
140
Intonation and Stress: Evidence from Hungarian
Example (21) is a declarative sentence that carries new information all the way through, (it has “broad focus”), that is why each lexical word, apart from the verb iszik ‘drinks’, receives major stress in it. The verb iszik, though it also conveys new information, is minor-stressed, as is normal when the F position before the verb is filled. 18 The F position here is filled by the word kávét ‘coffee-ACC’, which is a common noun without a determiner, and, as such, it is a kind of reduced complement (see É. Kiss 1981: 189–90). Reduced complements can be determinerless common nouns (22)a, verbal prefixes (22)b, and predicative adjectives (22)c:
(22) a. [kávét ]iszik, [moziba coffee-ACC drinks cinema-to [orvos ]lesz doctor will-be-3sg 19 b. [megjön, [elmegy PERF-comes away-goes 20 c. [szerencsés ]volt lucky was-3sg
]megy, goes
In non-emphatic declaratives, which carry broad focus, reduced complements are major-stressed and stand in F position before the minor-stressed verb, while other constituents in the sentence are major-stressed. Deaccenting the verb after a filled F is a case of syntactically motivated deaccentuation, effected by the application of rule (23).
(23) Post-Focal Verb Deaccentuation *
*
σ =obl: σ / S[. . . F V[(
σ0]. . . ] where V is a verb and F is a filled Focus. Domain: sentence. The reduction of the major stress to minor stress in function words is also the job of a postlexical deaccentuation rule, (24). The deaccentuation of function words is triggered by the syntactic category label (FW = function word) of these words. So deaccentuation here is a particular kind of syntactically motivated deaccentuation again.
(24) Function Word Deaccentuation (FWDA) * σ =obl: σ / S[. . . FW[( σ0]. . . ] where FW is a function word. Domain: sentence. Example (25)b illustrates the effect of FWDA.
Stress in Hungarian Words, Phrases and Sentences
(25)
a. [Meghívlak b. [Meghívlak
FW[[benneteket] FW[]benneteket]
141
[vacsorára. 3 FWDA [vacsorára.
Certain groups of lexical items have meanings that are akin to those of function words, and undergo the same deaccentuation processes as function words do. Therefore, we shall call them quasi function words. The first subtype of these contains so called “partial-degree constituents” (Varga 1983: 144). These are words expressing a degree somewhere between the positive and negative high or total degrees. To illustrate from English, the word some, for example, is a partial-degree constituent in English because it appears in the middle of the cline containing all (positive total), much/many (positive high), some (partial), little/few (negative high), none (negative total) see Table 6.2. Table 6.2 + total all
The degree cline (illustrated from English) + high much/many
partial some
– high little/few
– total none
Partial degree in Hungarian may appear in quantity: néhány ‘a few’, némi ‘some’, egy kicsi ‘a little’; temporal or spatial distribution: néha ‘sometimes’, néhol ‘at some places’, egyelore ‘for a while’; intensity: enyhén ‘slightly’, kissé ‘a little’, némileg ‘somewhat’, and in existentially quantified indefinite pronouns and adverbs: valaki ‘somebody’, valahol ‘somewhere’, valahogyan ‘somehow’, etc.:
(26) [Vettem ]néhány [nyakkendot. bought-1sg some necktie-ACC ‘I’ve bought a few ties.’ The two homophonous words valamennyi1 (total-degree: ‘all’) and valamennyi2 (partial-degree: ‘some’) are distinguished by their stressing in sentences like (27)a and (27)b:
(27)
a. [Idejött [valamennyi. here-came-3sg all ‘All of them have come here.’ b. [Idejött ]valamennyi. here-came-3sg some ‘Some of them have come here.’
142
Intonation and Stress: Evidence from Hungarian
Some semantically impoverished modifiers like bizonyos ‘certain’, valóságos ‘veritable’, szegény ‘poor’, kis ‘little’, etc. may also belong to quasi function words:
(28) [Láttuk ]szegény [Angéla ]nénit. saw-1pl poor Angela aunt-ACC. ‘We’ve seen poor Aunt Angela.’ The deaccentuation of quasi function words is also the job of FWDA, (24), with the added condition that it should apply to quasi function words as well. Further cases of syntactic deaccentuation include (i) words denoting measurement-units after numerals in attributive phrases (29)a, (29)b; (ii) appositives, titles and ranks standing before or after proper names (29)c, (29)d, (29)e; (iii) words modified by the negative particle nem ‘not’ or ne ‘don’t’ (29)f, (29)g; (iv) words modified by a question word (29)h:
(29) a.
[negyven 40
]kiló kilogram
[krumpli potato
‘40 kilos of potatoes’
b.
egy [kilencszáz ]lelkes a 900 soul-with ‘a village of 900 inhabitants’
[falu village
c.
[Angéla Angela
]néni aunt
‘Aunt Angela’
d.
[Rákóczi Rákóczi
]út Road
‘ Rákóczi Road’
e.
]doktor doctor
[Hidvégi Hidvégi
‘doctor Hidvégi’
f.
[nem not
szép nice
‘ It is not nice.’
g.
[ne don’t
menj go-IMP-2sg
‘ Don’t go.’
h.
[milyen how
]mély deep
‘ how deep’
Stress in Hungarian Words, Phrases and Sentences
143
When the focus position of a declarative sentence is filled by a constituent other than a reduced complement, the sentence is emphatic. In this case the constituent put into the focus position will take major stress (even when it is an otherwise non-major-stressed function word such as a personal pronoun), and the reduced complement, if there is one, will be banned from the focus position. The other major stresses will be reduced to minor stress (at least in the post-focal part of the sentence). 21 For instance, (30) is a non-emphatic sentence, having broad focus, with the reduced complement kávét in F position, and all the constituents in it are major-stressed. By contrast, the sentences in (31) have narrow focus and are, therefore, emphatic; the reduced complement (kávét) is outshifted and deaccented when it is in postverbal position, and either deaccented or accented when in pre-focal position. (30) [János [mindig [kávét ]iszik a [büfében. ‘John always drinks coffee in the snack-bar.’ (31) a
(i) ]János ]mindig a [büfében (ii) [János [mindig a [büfében ]kávét [kávét
]iszik ]iszik
]kávét. ]kávét.
b
(i) ]János (ii) [János
c
(i) ]Kávét ]János ]mindig a [büfében ]iszik. (ii) [Kávét [János [mindig a [büfében ]iszik. ‘It is in the snack-bar that John always drinks coffee.’
]mindig [mindig
a a
[büfében [büfében
]iszik. ]iszik.
All these cases of syntactic deaccentuation, together with (23) and (24), can be conflated into the following informal rule schema of Syntactic Deaccentuation: (32) Syntactic Deaccentuation (DA-1) * σ =obl: σ / S[. . . W[( σ0]. . . ] where W is (a) a verb after a filled focus position, see (23), (b) a function word or quasi function word, see (24), (c) a measurement-unit in an attributive phrase, (d) an appositive, title or rank associated with a proper name, (e) a word modified by a negative particle, (f) a word modified by a question word. (g) a reduced complement in postfocal position, within an emphatic sentence. Domain: sentence.
144
Intonation and Stress: Evidence from Hungarian
Although the rule schema is obligatory, in the special narrative style (Kálmán and Nádasdy 1994: 462) referred to above, its effect may be undone and the words involved may appear with major stresses, too. 6.6.2
Other kinds of deaccentuation
Deaccentuation may also occur for semantic and/or pragmatic reasons. In the following we shall consider some rather heterogeneous cases of deaccentuation, without attempting to be exhaustive. A great deal of deaccentuation is due to the fact that the constituent stands after an “eliminative stress”, (i.e. a strong contrastive stress; see Kálmán and Nádasdy 1994: 396), or is simply in the “informationally given” (i.e. presupposed) portion of the utterance. In (33)a and (33)b the verb tudjátok receives an eliminative major stress for emphasis and the subordinate clause after it reduces its own major stress to minor.
(33)
a. [Tudjátok, know-2pl b. [Tudjátok, know-2pl
hogy that ]hogy how
]dolgozom. ‘You know that I work.’ work-1sg dolgozom. ‘You know how I work.’ work-1sg Varga (1985: 209)
Another kind of deaccentuation is exemplified by the sentence pair in (34), displaying an aspectual difference. (34)a, where no deaccentuation takes place, has a ‘progressive’ interpretation: ‘At a certain point of time the activity described was in progress’. Example (34)b, with deaccented Pv constituents, conveys an ‘existential’ interpretation: ‘Until now the activity described has happened at least once’.
(34) a. [Mentem a [fel [lépcson. went-1sg up the stairs-on ‘I was just going up the stairs (when . . . )’ b. [Mentem ]fel a ]lépcson. ‘It has happened at least once until now that I went up the stairs.’ In (35)a the positive high degree quantifier sokan ‘many’ is major-stressed and its scope includes the negative particle even though the negative particle stands before it. In (35)b, however, the quantifier in the final Pv position is deaccented, and so its scope is included in the scope of the negative particle nem ‘not’.
Stress in Hungarian Words, Phrases and Sentences
145
(35) a. [Nem beszéltek [sokan. not talked-3pl many 22 ‘Many (of them) did not talk.’ b. [Nem beszéltek ]sokan. ‘Not many (of them) talked.’ In many cases deaccentuation is related to the fact that the deaccented constituent is being used within an inorganic block. (For the inorganic block, see section 5.3.2.). In (36)a the word természetesen ‘naturally’ is an adverbial of manner and is major-stressed, whereas in (36)b it is a sentence qualifier, that is a kind of inorganic block, and as such, it is deaccented:
(36)
[természetesen. a. [Mozogj move-IMP-2sg naturally b. [Mozogj, ]természetesen.
‘Move naturally.’ ‘Move, of course.’
Deaccentuation also takes place in ordinary yes–no questions, where the Pv constituents are deaccented even when they have not been presupposed, as in (37)a. By contrast, the primary stresses are preserved in so-called echo (surprised or repetitive) yes–no questions, as in (37)b, (consider also (6)c and (6)d in section 2.3.3). The words are: találkoztál ‘met-2sg’, az ‘the’ and édesanyámmal ‘mother-my-with’.
(37)
a. | 6Találkoztál az ]édesanyámmal? | ‘Did you meet my mother?’ b. | [Találkoztál az 6édesanyámmal? | ‘Is it really the case that you met my mother?’
Example (37)a can be uttered as an opening question in a conversation, that is when my mother has been neither mentioned nor seen. This situation is crucially different from English, where the Nuclear Stress Rule is at work in such sentences. The following is an attempt to sum up the various cases of semantic and/or pragmatic deaccentuation we have seen in a single schema, in the order of the examples (33)–(37):
(38) Semantic/Pragmatic Deaccentuation (DA-2) * σ =obl: σ / S[. . . W[( σ0]. . . ] where W is a word
146
Intonation and Stress: Evidence from Hungarian
(a) in the informationally given part of a sentence or after an eliminative stress, (b) in the Pv position of a sentence that has an existential aspectual interpretation, (c) meaning a positive high degree quantity in final Pv position, included in the scope of the negative particle nem, (d) within an inorganic block, (e) in the Pv position of a sentence that is used as an ordinary yes–no question. Domain: sentence.
Although more is to be said on major stress placement and deaccentuation in Hungarian phrases and sentences, we cannot deal with this issue any longer here. (For some further points see É. Kiss 1981, 1987; Kenesei 1998; Hunyadi 1999.)
6.6.3
Clash deletion: obligatory destressing
There is a postlexical rule called Clash Deletion (CD) (see Roca 1986: 355; and “Stress Deletion” in Halle and Vergnaud 1987: 238). This will delete the minor stress of a syllable which is immediately after a majoror minor-stressed syllable, within a string of words which begins with the left boundary of a maximal syntactic category and lasts until the left boundary of the next maximal syntactic category.
(39) Clash Deletion (CD) σ* σσ0 . . . Xmax[ ( =obl: 0 / Xmax[. . . ( σ where . . . does not contain Xmax[. Domain: string of words between any two consecutive left 23 boundaries of maximal syntactic categories.
Clash Deletion is a postlexical rule that affects the unbounded feet that were originally established in the lexical stratum, and carried over into the postlexical stratum of phonology. Consider (40), in which iszik, a verb after a filled F position, has been deaccented by (23) and, since it does not start a new maximal syntactic category (see (21)a), sört iszik is a domain for CD. As a result, it will undergo Clash Deletion and the original foot-head status of the syllable i- will be lost.
Stress in Hungarian Words, Phrases and Sentences
147
(40) János mindig sört iszik a büfében. John always beer-ACC drinks the snackbar-in ‘John always drinks beer in the snackbar.’ a. János mindig sört iszik a büfében. Q T b.
F
V
Pv
C
[János [mindig [sört ]iszik a [büfében. 3 [János [mindig [sört iszik a [büfében.
* * * * * (János (mindig (sört (iszik a (büfében. * * * DA: (János (mindig (sört (iszik a (büfében. * * * CD: (János (mindig (sört iszik a (büfében. Let us compare another pair of examples: (41) and (42). In (41) the compound [fott]csirke ‘boiled chicken’ satisfies the structural description of (39) and, consequently, Clash Deletion will delete the minor stress on the syllable csir-, yielding (41)b.
(41) Finom volt a fott csirke. nice was the boiled chicken ‘The boiled chicken was nice.’ Finom volt a fottcsirke. F V Pv [Finom ]volt a [fott]csirke. 3 [Finom ]volt a [fottcsirke. * * * (Finom (volt a (fott(csirke. * * DA: (Finom (volt a (fott(csirke. * * CD: (Finom (volt a (fottcsirke. By contrast, in (42) fott ‘was boiling’ and csirke ‘chicken’ belong to different maximal syntactic categories, and so the minor stress on csircannot be deleted; CD is not applicable.
148
Intonation and Stress: Evidence from Hungarian
(42) Fott csirke ebben a fazékban. the pot-in boiled-3sg chicken this ‘It has already happended at least once that chicken was boiled in this pot.’ Fott csirke ebben a fazékban. V Pv1 [Fott ]csirke
Pv2 ]ebben a ]fazékban.
* * (Fott (csirke
* * (ebben a (fazékban.
* DA: (Fott (csirke (ebben a (fazékban.
The relevant structural and prosodic differences are summarized in (43)a and (43)b.
(43) a. b.
6.7
N[[[fott][csirke]] S[ VP[ V′[ V[[fott] XP[]csirke]]]]
Summary
In this chapter we have embarked upon the topic of stressing in Hungarian. We found that the Hungarian stress system consists of major-, minor- and zero stress. We examined word stress in simple words, derivatives and ordinary compounds, and then gave a fairly detailed account of the stage of stress fixing in syntactic constructions, which takes place primarily in the form of deaccenting the major stresses that the words, inserted into the syntactic constructions, bring with them from the lexical phonology. In our rules we have made reference to syntactic domains such as words, major syntactic categories and the sentence, and so far to two kinds of prosodic domain: the syllable and the foot.
7 Rhythmical Variation in Phrasal Compounds
7.1
Introduction
In section 6.5.2 we introduced the notion of phrasal compounds and found that they differ from ordinary compounds in retaining more than one major stress. However, we did not examine the changes to which these major stresses can be submitted when the phrasal compound is preceded or followed by other major-stressed words in a phrase. These changes represent two kinds of rhythmical variation and it is these that we will concentrate on in the present chapter. In section 7.2 we present the two kinds of rhythmical variation that phrasal compounds may undergo. Section 7.3 shows that analyses based on the intensity differences of major stresses cannot account for the facts of rhythmical variation in Hungarian, and therefore a “split analysis” is proposed in section 7.4. Finally, in section 7.5, we draw the most important conclusion of the discussion: that the intensity differences of major stresses are irrelevant for the treatment of rhythmical variation. This confirms what we have asserted throughout the book and especially in Chapter 6: it is the presence or absence of major stresses that matters, and not the degree of their physical strength. The chapter is a shortened and slightly revised version of Varga (1998).
7.2
The facts of rhythmical variation in Hungarian
7.2.1 Phrasal compounds and the two kinds of rhythmical variation As we saw in section 6.5.2, phrasal compounds are lexical items that have exceptional, phrase-like stressing: they have more than one major stress, one on the first syllable of each of their stressable constituents, 149
150
Intonation and Stress: Evidence from Hungarian
for example [ütött-[kopott ‘battered’, [kétszáz[kilenc ‘209’ (in which kétszáz = ‘200’ and kilenc = ‘nine’ ). For a start, we shall assume that the first of the two major stresses in such compounds is phonetically stronger than the second. Some, but not all, of the phrasal compounds are rhythmically variable (RV) words, for example as in (1)a. An RV word is double-accented, that is it has two major-stressed syllables, and is capable of undergoing two kinds of rhythmical variation when it is embedded within syntactic phrases or larger numeric compounds. For instance, (1)a shows the isolated pronunciation of the RV word tizenegy ‘11’, in which the word has the stress pattern of (stronger) major + (weaker) major stress. This may become non-major + major, as in (1)b, and major + non-major, as in (1)c. In the examples the line of text follows the schematic pitch curve which the examples would take in declarative utterances.
(1) a.
tizenegy ‘11’ (literally: ‘one on ten’) Pattern:
[[tizen[egy] [[B
[C ] ti
b.
zenegy
pont tizenegy ‘exactly 11’ Pattern:
[[pont [tizen[egy] ] [[A
[B
pont
c.
(Trochaic Reversal)
[C ] ]
e tizen g
y
tizenegy pont ‘11 points’ Pattern:
[ [[tizen]egy] [pont] [ [[B ti
C
] [D
(Trochaic Confirmation)
]
p zenegy on t
[B C] stands for the two, potentially major-stressed elements of the word [tizen[egy. The syntactic pattern is right-branching [A [B C]] in (1)b and left-branching [[B C] D] in (1)c. In both types of context one of the
Rhythmical Variation in Phrasal Compounds 151
major stresses of the phrasal compound is deaccented, that is reduced to minor stress. This minor stress is further reduced to zero stress by Clash Deletion (see section 6.6.3) in (1)b, but not in (1)c. The more radical type, occurring in the right-branching pattern, (1)b, can be called Trochaic Reversal. This name encapsulates the assumption that in the original sequence of two major stresses the first one is phonetically stronger than the second, and this strong + weak (i.e. “trochaic”) prominence is reversed and becomes weak + strong by deaccenting the initial major stress – the one on B – thus producing words with a single, word-final accent: ]tizen[egy. The result is in dramatic contrast to the general Hungarian tendency of having a single, word-initial accent. Trochaic Reversal occurs only in RV words.1 The less radical type, shown in the left-branching pattern of (1)c, will be called Trochaic Confirmation, because the final element of the phrasal compound, that is C, which is assumed to be phonetically somewhat weaker than the first, becomes even weaker by deaccenting its major stress, and so the trochaic pattern becomes even more conspicuous. 2 This change brings the double-accented word into line with the general Hungarian tendency of having a single, word-initial accent: [tizen]egy. Trochaic Confirmation can occur not only in RV words but also in nonRV phrasal compounds when they are followed by a major-stressed word in the same phrase. For instance, [ütött-[kopott ‘battered’ in (2)a can be realized as [ütött-]kopott, see (2)c, but not as * ]ütött-[kopott, see (2)b.
(2) a.
ütött-kopott ‘battered’ (literally: beaten-worn) [ütött-[kopott
b.
nagyon ütött-kopott ‘very battered’ *
c.
[nagyon
]ütött-[kopott
(Trochaic Reversal)
ütött-kopott villamosok ‘battered trams’ [ütött-]kopott
[villamosok
(Trochaic Confirmation)
So Trochaic Reversal and Trochaic Confirmation are asymmetrical processes in Hungarian. While the former is restricted to RV words, the latter can affect all phrasal compounds (PCs), RV and non-RV alike. This is shown in Table 7.1.
152
Intonation and Stress: Evidence from Hungarian
Table 7.1
The asymmetry of Trochaic Reversal and Trochaic Confirmation
Trochaic Reversal Trochaic Confirmation
RV-PC
Non-RV-PC
+ +
– +
Trochaic Confirmation makes the originally double-accented words similar to the overwhelming majority of Hungarian words: it leaves only the initial accent on them. This is why frequently used forms showing Trochaic Confirmation may be lexicalized. For many speakers, isolated RV words may have two parallel forms, a double-accented form capable of rhythmical variation: [tizen[egy ‘11’, and an initial-accented form which is the lexicalized result of Trochaic Confirmation: [tizen]egy. Such speakers may use either form as a reply to the question Hány óra van? ‘What time is it?’, see (3).
(3) A:
-
Hány óra van?
B:
-
[Tizen[egy. / [Tizen]egy.
That the double-accented form is a reality even for these speakers is evidenced by the fact that the same speakers will apply Trochaic Reversal in, for example, (4), which presupposes the double-accented form as the underlying representation.
(3) Fél half
tizenegy. 11
‘Half past 11’ [Fél tizen[egy. Rhythmical variation in Hungarian (at least the more radical kind that we have called Trochaic Reversal), is severely constrained and is restricted to the following classes of double-accented compounds:
(5) Rhythmically variable words (RV words) (i) Inherently double-accented numeric compounds: (a) Two-digit cardinal and ordinal numerals in which the second digit is not zero, e.g. [tizen[három ‘13’ (in which tizen = ‘on 10’, literally: ‘10-on’, and három = ‘three’) , [nyolcvan[hat ‘86’ (in which nyolcvan = ‘80’, hat = ‘six’); [negyven[ötödik ‘45th’ (in which negyven = ‘40’, ötödik = ‘fifth’).
Rhythmical Variation in Phrasal Compounds 153
(b) Three-digit cardinal or ordinal numerals in which either the second or the third digit is zero, e.g. [ötszáz[három ‘503’ (in which ötszáz = ‘500’, három = ‘three’) or [ötszáz[harminc ‘530’ (in which harminc = ‘30’); [száz[tizedik ‘110th’ (in which száz = 3 ‘100’, tizedik = ‘tenth’). (c) Four-digit cardinal or ordinal numerals in which two of the last three digits are zero, e.g. [ezer[kilencszáz ‘1900’ (in which kilencszáz = ‘900’), [ezer[kilencven ‘1090’ (in which kilencven = ‘90’), [ ezer [ kilenc ‘1009’ (in which kilenc = ‘nine’); [ezer[ötszázadik ‘1500th’ (in which ötszázadik = ‘500th’). (ii) Derivatively double-accented numeric compounds: These contain three accents underlyingly, but retain only two accents after Trochaic Reversal takes place on an internal constituent in them; e.g. if the constituent tizenhat ‘16’ within [kétszáz[tizen[hat ‘216’ undergoes Trochaic Reversal, the originally triple-accented word becomes [kétszáz]tizen[hat, which is a 4 derivatively double-accented numeric compound. (iii) Ugyan-compounds: These are composed of the bound stem ugyan- ‘the same...’ and a demonstrative stem, e.g. [ugyan[annyi ‘the same amount’ (in which annyi = ‘that much’), [ugyan[az ‘the same’ (in which az = ‘that=DEM’), [ugyan[olyan ‘exactly like that’, [ugyan[ott ‘in the same place’ (in which ott = ‘there’), [ugyan[akkor ‘at the same time’ (in which akkor = ‘then’), etc. Other double-accented compounds, e.g. [ütött-[kopott ‘battered’, etc. are not RV, but Trochaic Confirmation may also affect them. 5 7.2.2
The number of syllables between the accents
Rhythmical variation in the stressing of Hungarian is never strictly obligatory but it is very likely when a major-stressed syllable of an RV word is flanked by adjacent major-stressed syllables on both sides. The reduction of the medial major stress in this case will leave one unaccented syllable, as is shown in the righthand versions of examples (6) and (7):
(6) pont
száztíz
exactly
100-ten
‘exactly
110’
[pont [száz[tíz [[A
[[B
[C ] ]
(Trochaic Reversal)
3
[pont száz[tíz
154
Intonation and Stress: Evidence from Hungarian
(7) száztíz
kérdés
100-ten
(Trochaic Confirmation)
question
‘110 questions’ [száz[tíz [kérdés [ [B
C]
3
[száztíz [kérdés
D]
The process is also quite likely when the number of interaccentual syllables created by rhythmical variation is two, as in examples (8) and (9):
(8) Pattern:
[A [B C] ];
(Trochaic Reversal)
a. százharmincnégy ‘134’ 100-30-
four
[száz[harminc[négy b. fél half
tizenketto
3
[százharminc[négy
‘half past 11’;
lit. : ‘half 12’
ten-on-two
[fél [tizen[ketto 3 [fél tizen[ketto c. négy huszonötkor four-20-
‘at 4.25’
five-at
[négy [huszon[ötkor d. pont
3
ugyanaz 'exactly the same'
exactly
same
[pont [ugyan[az (9) Pattern:
3
30-
four
b. huszonöt five
‘34 books’
book
ötvenért
3
[harminc]négy [könyv
‘for a price of 25.50’
50-for
[huszon[öt [ötvenért c. ugyanaz
(Trochaic Confirmation)
könyv
[harminc[négy [könyv
same
[pont ugyan[az
[ [ B C ] D ];
a. harmincnégy
20-
[négy huszon[ötkor
tökben
3
[huszon]öt [ötvenért
‘it’s just the same’
marrow+in
[ugyan[az [tökben
3
[ugyan]az [tökben
Rhythmical Variation in Phrasal Compounds 155
Rhythmical variation is also possible in Hungarian when it leaves a sequence of three or four syllables between the accents, though with four interaccentual syllables it may occur less frequently.
(10) Pattern:
[A [B C] ];
(Trochaic Reversal)
a. négyszázötvenketto 400-
50-
‘452’
two
[négyszáz[ötven[ketto b. negyed
tizenketto
quarter
10-on-two
‘a quarter past 11’;
3
lit.: ‘quarter 12’
[negyed [tizen[ketto c. mindig
ugyanaz
always
3
Fiat
[negyed ]tizen[ketto
‘always the same’
same
[mindig [ugyan[az d. Fiat
[négyszáz]ötven[ketto
3
ezerötszáz
[mindig ]ugyan[az
‘Fiat 1500’
1000-500
[Fiat [ezer[ötszáz e. október October
3
huszonharmadika
[Fiat ]ezer[ötszáz ‘October 23rd’
20-on third
[október [huszon[harmadika 3 [október ]huszon[harmadika (Trochaic Confirmation)
(11) Pattern: [ [ B C ] D ]; a. a the
tizenhárom
vértanú
ten-on-three
martyr
a [tizen[három [vértanú b. ugyanaz same
a
kérdés
the
question
[ugyan[az a [kérdés c. ezerötszáz 1000-500
‘the 13 martyrs’ 3
a [tizen]három [vértanú ‘the same question’
3
vagonnal
[ugyan]az a [kérdés ‘with 1500 waggons’
waggon-with
[ezer[ötszáz [vagonnal
3
[ezer]ötszáz [vagonnal
156
Intonation and Stress: Evidence from Hungarian
d. a the
huszonharmadik
nap ‘the 23rd day’
20-
day
third
a [huszon[harmadik [nap
3 a [huszon]harmadik [nap
There is a limit to the number of unaccented syllables that can occur between two accents as a result of rhythmical variation. This is expressed in (12):
(12) Five-Syllable Constraint If the application of rhythmical variation results in a string of more than four unaccented syllables between the two remaining accents, then rhythmical variation becomes unlikely. For instance, rhythmical variation is unlikely in (13) or (14), where the number of interaccentual syllables is too high:
(13) Pattern:
[A [B C]];
(Trochaic Reversal )
kilencezer kilencszáztíz 9000-
900-
ten
[kilencezer[kilencszáz[tíz (14) Pattern:
‘9910’ 3
?[kilencezer]kilencszáz[tíz
[[B C] D];
(Trochaic Confirmation)
kilencezer kilencszáz
utas
9000-
passenger
900
‘9900 passengers’
[kilencezer[kilencszáz [utas 3 ?[kilencezer]kilencszáz [utas Some phrases are concatenations of two RV words: [[B1 C1] [B2 C2]]. These will be submitted to both Trochaic Confirmation and Trochaic Reversal. In such phrases either or both of the internal major stresses may be lost, and there is no limit on the number of interaccentual syllables between the remaining major stresses.
(15) Pattern: [[B1 C1] [B2 C2]] a. tizenhárom negyvenöt ‘13.45’ ten-on-three 40five [tizen[három-[negyven[öt 3 [tizen]három-[negyven[öt or: [tizen[három-]negyven[öt or: [tizen]három-]negyven[öt
Rhythmical Variation in Phrasal Compounds 157
b. ezerkilencszázötvenhat ‘1956’ 1000-90050- six [ezer[kilencszáz[ötven[hat 3 [ezer]kilencszáz[ötven[hat or: [ezer[kilencszáz]ötven[hat or: 6 [ezer]kilencszáz]ötven[hat 7.2.3
Parallel phenomena in English
Rhythmical variation in Hungarian is reminiscent of the parallel phenomenon of rhythmical variation in English, in which, for example, the isolated form [thir[teen of (16)a may become [thirteen in (16)b, and thir[teen in (16)c. (The symbol [[] in the English examples also represents major stress, but the rightmost major stress is definitely stronger than the one before it because it is associated with the initiation of a nuclear intonation contour.)
(16)
a. [thir[teen b. [thirteen [men c. [just thir[teen
(Iambic Reversal) (Iambic Confirmation)
The processes shown in (16)b and (16)c are the two kinds of rhythmical variation in English. The more radical kind shown in (16)b is often called Iambic Reversal (see, for examples Liberman and Prince 1977), because the originally weak + strong (i.e. “iambic”) prominence pattern of a word is reversed into a strong + weak prominence pattern when the word is followed by an initially-accented word in the same phrase. Iambic Confirmation is the name I use for the less radical rhythmical change (16)c: the already weaker initial accent becomes even weaker.7 Rhythmical variation in Hungarian differs from its counterpart in English in two important respects. First, Hungarian rhythmical variation is lexically constrained, whereas the parallel variation in English is not restricted to certain classes of words; apart from a few lexical exceptions, practically all words and phrases that have the underlying minor + major stress pattern and are embedded in larger phrases may systematically undergo the variation; consider, for example, [pontoon [bridge, [Dundee [marmalade, [academic [discipline, [good-looking [tutor; [three red [shirts. Secondly, the kind of asymmetry we find between the two processes of rhythmical variation in Hungarian does not exist in English; the same set of words is targeted by both types of variation.
158
Intonation and Stress: Evidence from Hungarian
In Hungarian, Trochaic Confirmation may affect all phrasal compounds but Trochaic Reversal may affect only a subset of phrasal compounds, viz. RV words.
7.3
The analysis of rhythmical variation in Hungarian
7.3.1
Analyses based on different accentual strengths
Varga (1998) examined the possibility of adapting Selkirk’s (1984) gridbased model, and Hayes’ (1984) metrical adjunction theory (together with its extensions in Kager and Visch 1988 and Hayes 1995) to data on Hungarian rhythmical variation, and found that such adaptations raised serious problems. The most important single problem in these adaptations was that they could indicate and manipulate subtle prominence differences between stresses, which, however, turned out to be irrelevant to Hungarian rhythmical variation, while they were unable to express the non-relational aspect of Hungarian major stresses, which alone was relevant to Hungarian rhythmical variation. 7.3.1.1
The adaptation of Selkirk’s model
In this section we shall build grids for Hungarian words on the basis of Selkirk’s (1984) universal Text-to-Grid Alignment Rules. We assume that, as a result of these rules, all Hungarian lexical words will come out of the lexicon with a column of four grid marks (beats) over their first (i.e. major-stressed) syllable, and with columns of two grid marks over the unstressed syllables, as in (17)a. 8 In compound words, the Compound Stress Rule (CSR) will eliminate the Level 4 (L4) beat over the first syllable of a non-initial immediate constituent, leaving a minor stress on that syllable, see (17)b.
(17) a. L4: L3: L2: L1:
x x x xx x xx [iskola ‘school’
b. x x x x xx x x xx x x [iskola]táska ‘schoolbag’
In phrases, the Hungarian version of the Nuclear Stress Rule (NSR) will promote the prominence of the leftmost constituent in the phrase in such a way that it should become stronger than the strongest prominence on its right, and grid construction will proceed in a cyclical fashion so that Level 5 (L5) or higher levels will be reached, see (18)a.
Rhythmical Variation in Phrasal Compounds 159
Phrasal compounds in Hungarian, including RV words, have a peculiar status: they are words, but with regard to their stressing they are treated as if they were phrases and their constituents as lexical words. The rule responsible for their stressing is the Nuclear Stress Rule, which, in this context, will be called the Phrasal Compound Stress Rule (PCSR). The underlying grids for ütött-kopott ‘battered’ and tizenhárom ‘13’ can be constructed as shown in (18)b and (18)c, respectively.
(18) a. NSR
x x x x x x x [édes
x x x x x x [álom ‘sweet dream’
b. PCSR (non-RV)
x x x x x x x x x x x x x [ütött-[kopott
c. PCSR (RV)
x x x x x xx xx xx xx [tizen[három
By applying the Text-to-Grid Alignment rules that we have described, we can produce underlying metrical grids for Hungarian words and phrases. When a syllable is associated with a beat on L4, or higher, that syllable is major-stressed and will be pitch accented. A major-stressed syllable minimally requires a beat on L4. When the highest beat associated with a syllable is on L3, the syllable is minor-stressed. And when the highest beat associated with a syllable is on L2, the syllable is unstressed. The underlying grids may be modified by optional Grid Euphony Rules, whenever their structural description is satisfied in a cycle. These Grid Euphony Rules are the Hungarian versions of Beat Addition, Beat Deletion
160
Intonation and Stress: Evidence from Hungarian
and Beat Movement. Each of them is the mirror image of its English counterpart: the English and Hungarian versions of these rules differ in their parametric settings. A general condition for all Hungarian Grid Euphony Rules is that the bottom level in these rules is minimally L2.
(19) a. Hungarian Beat Addition (BA) x xx 3 x x (Right dominant, L–R.) b. Hungarian Beat Deletion (BD) x x xx x xx x x 3 c. Hungarian Beat Movement (BM) x x xx x x xxx 3 x x x Restriction: BM can only affect an RV word if the clash is within a constituent which immediately dominates the RV word. It is these rules that are responsible for Trochaic Reversal and Trochaic Confirmation in the RV words of Hungarian. Since they affect L4 grid marks (i.e. major stresses) in RV words only, they differ from the Selkirkian rules in an important respect: they need information from outside the grid, viz. they have to ‘see’ whether or not a word is an RV lexical item:
(20) General Restriction on Hungarian Grid Euphony Rules The affected (i.e. added, deleted, or moved) beat cannot be on L4, unless it is over a syllable of an RV word. Trochaic Confirmation requires the application of both Beat Addition and Beat Deletion, as is shown in (21):
(21) a. Lexicon
x x x x x x xx xx x xx xx x [tizen-, [három, [pont
Rhythmical Variation in Phrasal Compounds 161
b. Cycle 1 PCSR
as in (18)c
c. Cycle 2 NSR
x x x x x x x xx xx x xx xx x [tizen[három [pont [[B C ]D ]
d. BA, TPPC
x x x x x x x x x xx xx x xx xx x [tizen[három [pont
e. BD
x x x x x x x xx xx x xx xx x [tizen]három [pont ‘13 points’
The TPPC in (21)d is Selkirk’s Textual Prominence Preservation Condition (Selkirk 1984: 56, 100–1).9 However, Beat Deletion is not permitted in (22)b, where the L4 beat which it would delete (over the syllable or-) is not in an RV word. Example (22)b is the result of NSR for the first cycle (egy magyar orvos), NSR for the second cycle (egy magyar orvos sikere), BA (twice) and the TPPC. (To save space, numbers will replace the columns of grid marks over the syllables, each number corresponding to the number of grid marks in the column.)
(22) a. BA (twice), TPPC
2 egy
6 2 4 2 [magyar [orvos
52 3 [sike]re
162
Intonation and Stress: Evidence from Hungarian
b. BD impossible
2 6 2 3 2 52 3 * egy [magyar ]orvos [sike]re ‘the success of a Hungarian physician’ (22)b is impossible because of the restriction in (20).
Unfortunately, however, Beat Deletion, as formulated in (19)b, cannot eliminate the second accent of a non-RV phrasal compound (e.g. [ütött[kopott ‘battered’) either, because it is not an RV word, although such phrasal compounds may lose their second accent before an accented word.
(23) BD impossible BD
x x x x x x x x x x x x x x egy [ütött -]kopott ‘a battered car’
x x x x x x [autó
Example (23) is impossible because of (20), although it would be correct. Nevertheless, we cannot reformulate the Beat Deletion rule so as to enable it to handle cases like (23) because the rule is also needed in Trochaic Reversal (see below), where non-RV phrasal compounds like [ütött- [kopott never lose their accent. Trochaic Reversal will require the application of Beat Movement and Beat Deletion, see (24).
(24) a. Lexicon
x x x x x x x x xx xx x x xx xx [mind, [tizen-, [három a, ----------------------------------------------b. Cycle 1 PCSR as in (18c) -----------------------------------------------
Rhythmical Variation in Phrasal Compounds 163
c. Cycle 2 NSR
x x x x x x x x x x x x x x x x x x x x x [mind a [tizen[három [A [ B C ]]
d. BM
x x x x x x x x x x x x x x x x x x x x x [mind a [tizen[három
e. BD
x x x x x x x x x x x x x x x x x x x x [mind a ]tizen[három
‘all the thirteen’
But Beat Movement cannot take place in (25)b because it would move an L4 beat in a non-RV word. Example (25)b is the result of CSR and NSR.
(25)
a.
NSR
b.
BM impossible
2 5 4 22 3 2 egy [új [iskola]táska
2 5 3 22 4 2 * egy [új ]iskola[táska ‘a new schoolbag’ (25)b is impossible because of the restriction in (20). Hungarian Beat Movement has a special restriction (given in 19)c, viz. that it is only applicable to an RV word if the clash that it resolves is within a constituent that immediately dominates the RV word. Consequently, Beat Movement will not occur in (26)f:
164
Intonation and Stress: Evidence from Hungarian
(26) a. Lexicon
x x x x x x x x x xx x x x x x x xx x x a [hajó, [tizen-, [három,
x x x xx x xx [utasa
-------------------------------------
b. Cycle 1 PCSR as in (18c)
--------------------------------------------c. Cycle 2 NSR
d. BA, TPPC
e. BD
x x x x x x x x x x x x x [tizen[három
x x x x x x x x [uta]sa
x x x x x x x x x x x x x x [tizen[három
x x x x x x x x x x [uta]sa
x x x x x x x x x x x x x x x x x x x x x x x [tizen]három [uta]sa ---------------------------------------------
Rhythmical Variation in Phrasal Compounds 165
f.
Cycle 3 NSR
x x x x x x x x x x x
x x x x xx xx
x x x x x
x x x x xxx xxx
a [hajó [tizen]három [uta]sa [ E
[[ B
C
] D
]]
‘the ship’s 13 passengers’ Although there is a clash in (26)f, it cannot be resolved by Beat Movement because the clash is not within a constituent that immediately dominates the RV word tizenhárom. If the Beat Movement rule did not have this restriction and would be carried out in (26)f, its application would eventually lead to the wrong pattern * a [hajó ]tizen[három [utasa (through BM, BD, BM). To sum up, we have found that the adaptation of Selkirk’s grid-only model (1984) to Hungarian rhythmical variation faces a number of problems. First, the Hungarian versions of the Grid Euphony Rules have to be lexically constrained, which is prohibited in Selkirk’s original framework. As shown in the examples, Hungarian Beat Addition, Beat Deletion and Beat Movement cannot add, delete or move L4 grid marks (i.e. primary stresses) unless these grid marks are within RV words. Secondly, as (26)f has shown, the Hungarian version of Beat Movement can only affect an RV word if the clash that it resolves is within a constituent that immediately dominates the RV word. Again, such syntactic stipulations are excluded from Selkirk’s framework. Thirdly, though the analysis accounts for both kinds of rhythmical variation in RV words (i.e. Trochaic Reversal and Trochaic Confirmation), it ignores those cases of Trochaic Confirmation which occur in double-accented non-RV words, see (23). That is to say, it treats Trochaic Reversal and Trochaic Confirmation as if they were symmetrical, but is still unable to provide a unified account for them. Fourthly, in the Hungarian version of Selkirk’s model the levels of the grid have categorical, non-relative meaning: L4 and higher grid marks show pitch accent (major stress). However, since the analysis has a new step for every prominence change, including the insignificant changes in prominence of major stresses, it unnecessarily differentiates the pitch accents according to their strengths, see, for example, the different realizations of [mind a [tizen[három in (24)c and (24)d.
166
Intonation and Stress: Evidence from Hungarian
7.3.1.2
The adaptation of metrical adjunction theory
In Hayes’ metrical adjunction theory, both kinds of English rhythmical variation, that is both Iambic Reversal and Iambic Confirmation, were collapsed into a single adjunction rule called Rhythmic Adjustment (Hayes 1984). This has been extended to rhythmical variation in Dutch by Kager and Visch (1988). Rhythmic Adjustment in Dutch occurs not only in a leftward direction, as in English, but also in a rightward direction, as in Hungarian. Kager and Visch (1988: 52) define rightward Rhythmic Adjustment as is shown in (27):
(27)
Rhythmic Adjustment (Rightward) In the configuration ...DTE...Y X..., adjoin Y to X.
This is the mirror image of leftward Rhythmic Adjustment, which Hayes (1984) had set up for English. “Adjunction” here means that the sequence YX is made a constituent, with Y being weak relative to X. Hungarian is characterized by rightward Rhythmic Adjustment, as given in (27). The phrases (28–31) below exemplify rightward Rhythmic Adjustment in Dutch and Hungarian. Each example contains the metrical tree, the grid, and also the schematic intonation curve with underlining below the stressed syllables. In the first Dutch example, (28), taken from Kager and Visch (1988: 53), the compound voetbalveld (in which voet = ‘foot’, bal = ‘ball’, veld = ‘field’) means ‘soccer field’. The rule functions as Trochaic Confirmation on voetbal, and it carries out a rebracketing. As a result, bal, the final constituent of voetbal feels weaker than before the adjustment.
(28)
x x x x x x x x x x x x x 3 x x x [voet bal]veld voet[balveld] DTE Y X DTE Y X s w w s w s s voet bal veld
w 3
voet bal veld
Rhythmical Variation in Phrasal Compounds 167
In the next Dutch example, (29), the compound zaalvoetbal (in which zaal = ‘gym’) means ‘indoor soccer’ (Kager and Visch 1988: 53). The rule functions as Trochaic Reversal on voetbal. In this case the adjunction is vacuous in terms of constituent structure but renders Y weak in relation to X; it carries out a relabelling.
x x x x
(29)
x x x x x
3
x x x x
x x x x x
zaal [voetbal]
zaal [voetbal]
DTE
Y X
DTE
Y X
s
s w
s
w s
w z
a a l
w zaal 3
voetbal
voetbal
The problem with the tree-plus-grid representations produced in metrical adjunction theory is that they are able to show only one pitch accent reliably, viz. the one on the syllable with the highest column in the grid (i.e. the DTE). In Hayes’s view “phrasal stress is an independent domain, and pitch accents are constrained to attach to the strongest available stresses” (Hayes 1995: 370). What is not clear is how many of the strongest stresses should be associated with pitch accents: that is whether or not the syllables with the second, third, etc. highest column are also pitch-accented. If we apply metrical adjunction theory and rightward Rhythmic Adjustment to Hungarian examples like száztíz pont ‘110 points’ or pont száztíz ‘exactly 110’, we shall get the same trees and grids for Hungarian as for Dutch voetbalveld and zaalvoetbal, respectively. But despite the representational identity of the corresponding Hungarian and Dutch examples, the pre-adjustment versions of the Hungarian examples contain three pitch accents while their Dutch parallels have only one, and the post-adjustment versions of the Hungarian examples contain two pitch accents, whereas the parallel examples in Dutch have only one.
168
Intonation and Stress: Evidence from Hungarian
(30) Cf. (28) x x x x x
x x
3
x x x x
x x
x x x
[száz[tíz pont]
[száztíz] pont DTE Y
X
DTE
Y X
s w
w
s
w s
s
w száz
sz á t zí
3
p
o
z
tíz
p
o
n
t
n
t
(31) Cf. (29) x x x x
x x x x x
3
x x x x
x x x x x
pont [száztíz]
pont [száztíz]
DTE
Y X
DTE
Y X
s
s w
s
w s
w p
o
w pont
n
sz t
3
á t z í
z
t száz í
z
The s/w labels in the trees and the column heights in the grids offered by metrical adjunction theory can indicate prominence relations, but cannot express the primarily non-relational aspect of Hungarian major stresses. Hungarian major stresses are pitch accents and their majorstressed status is unaffected by their intensity differences. 7.3.2
Gussenhoven’s pitch accent-based analysis
Hungarian rhythmical variation is based on the existence and deletability of pitch accents. However, the mechanisms of the grid-only model or
Rhythmical Variation in Phrasal Compounds 169
the tree-plus-grid model of metrical phonology are unable to show these pitch accents or the manipulations affecting them, because they are concerned solely with the relative strength of stresses. Therefore an alternative framework is needed, whose adaptation to Hungarian can adequately handle pitch accents without differentiating them according to their relative phonetic strengths. Such a framework is offered by Gussenhoven (1991). Examining rhythmical variation in English, Gussenhoven concludes that both kinds of rhythmical variation can be regarded as deaccentuation of the medial accent in a series of three accents within the same phonological phrase. The Hungarian version of Gussenhoven’s Rhythm Rule (originally proposed for English) is given in (32):
(32) Hungarian Rhythm Rule (based on Gussenhoven 1991) * =opt: 0 / XP[* * ...]XP The rule works from left to right. Conditions: XP is a maximal syntactic category (a) which immediately dominates the RV word to which the accent to be deleted belongs, or (b) which immediately dominates, and starts with, a non-RV phrasal compound to which the accent to be deleted belongs. Gussenhoven’s accent-based approach is better suited to Hungarian than other existing models because it does not distinguish accentual strengths and, consequently, does not require metrical grids.10 However, when trying to account for Trochaic Confirmation in non-RV phrasal compounds (consider [ütött-[kopott), and thus to incorporate the asymmetry of Hungarian rhythmical variation, we have to add the ad hoc condition (b) to the Rhythm Rule in (32). Condition (a) is symmetrical because it accounts for both kinds of variation in RV phrasal compounds, and it is immaterial whether the RV compound is initial or non-initial in the phrase. Condition (b), on the other hand, is asymmetrical, accounting only for Trochaic Confirmation in non-RV phrasal compounds, and these compounds must be initial in the phrase. Another problem is that the adapted version of Gussenhoven’s model has to make reference to lexical class and syntactic constituency because rhythmical variation in Hungarian is confined to certain classes of lexical items and to certain kinds of syntactic constituents. This is
170
Intonation and Stress: Evidence from Hungarian
not permitted in Gussenhoven’s original model, which refers purely to a prosodic domain (the phonological phrase). These considerations have led to the emergence of the Split Analysis of Hungarian rhythmical variation (see Varga 1998). This analysis preserves Gussenhoven’s notion of accent, and claims that Trochaic Reversal belongs to the precompiled lexical rules of Hayes (1990), while Trochaic Confirmation is postlexical, belonging to the P1 rules of Kaisse (1985, 1990). 7.3.3
The split analysis
For Kaisse (1985, 1990), P1 rules (“external sandhi rules”) are a special subset of postlexical rules, which are “closest to the lexicon and thus share many characteristics with lexical rules” (Kaisse 1990: 130) and which “apply after all lexical rules and before all P2 rules” (Kaisse 1990: 128), the latter being the classical postlexical rules. Kaisse (1990: 135–7) proves fairly convincingly that English rhythmical variation has a number of P1 (i.e. lexical-rule-like) characteristics; for example it is sensitive to nested compound and syntactic bracketing. The possibility of explaining Hungarian rhythmical variation in terms of P1 rules is somewhat less clear, but acceptable as a first approximation (see Varga 1998: 257–8). Hayes (1990) says that Kaisse’s P1 rules are in fact lexical rules, more precisely precompiled lexical rules. This is how Hayes (1990: 87) introduces the notion of precompiled rules: all phonological rules fall into two classes: (a) truly phrasal rules, which apply postsyntactically and may refer only to the levels of prosodic hierarchy; (b) lexical rules, which apply presyntactically within the lexicon. Of the latter, a subset are “precompiled rules”, which derive multiple diacritically-marked allomorphs for certain classes of words. At the interface of syntax and phrasal phonology the appropriate diacritically marked allomorphs are inserted in the relevant syntactic contexts. The paradox that syntactic rules may not refer to phonological properties but may refer to properties of individual words and morphemes has led Hayes to think that words appear in syntactic representations as abstract place markers that lack phonological content. Lexical insertion then takes place in two steps: the first is insertion of the abstract place markers into the syntactic trees, the second is phonological instantiation of the abstract place markers with phonemic material. Consequently, most of the phonological rules that have been analysed in the past as
Rhythmical Variation in Phrasal Compounds 171
making reference to syntactic representations (so called “direct-syntax rules”) should be seen as selecting the appropriate precompiled allomorphs for phonological instantiation: “phrasal allomorphs may be derived by phonological rule within the lexical phonology, so that whole classes of words will have multiple precompiled allomorphs” (Hayes 1990: 92). Hayes (1990: 103) also claims that his precompiled rules correspond to Kaisse’s P1 rules and that “the simplest explanation for why P1 rules act as if they were lexical is to assume that they are lexical”, that is precompiled. However, Kaisse does not put an equation mark between P1 rules and precompiled rules. She says that “[p]recompiled rules might be partly diagnosable by their having lost even more phonetic motivation than P1 rules” (Kaisse 1990: 130). It seems that this loss of even more phonetic motivation is exactly what characterizes Hungarian Trochaic Reversal. In Hungarian, rhythmical variation can take place only in doubleaccented phrasal compounds, and such words are in a minority. Trochaic Confirmation may affect all these double-accented phrasal compounds, that is not only RV words like [tizen[egy ‘11’ but also nonRV phrasal compounds like [ütött-[kopott ‘battered’. However, Trochaic Reversal applies only to a subset of these phrasal compounds, viz. to RV words like [tizen[egy. So Trochaic Reversal is even more restricted than Trochaic Confirmation. There is no way to account for this asymmetry other than that it is a lexical property of RV words to accept Trochaic Reversal, and it is a lexical property of the other phrasal compounds to resist it. Therefore, while a P1 (postlexical) analysis seems plausible for Trochaic Confirmation, a precompilation (lexical) analysis seems more suitable for Trochaic Reversal. 11 Adapting Hayes’s precompilation theory to Hungarian Trochaic Reversal, we shall use the diacritic BC for RV words that are to be inserted in the phonological instantiation frame of Trochaic Reversal. The letters B and C represent the two accentable components. An asterisk over a capital letter represents the pitch accent on the word (or constituent) which the capital letter stands for. The lexicon of Hungarian includes the following “phonological instantiation frame”:
(33) Trochaic Reversal Frame * A / XP[ ]RV]XP RV[ * * RV[B C]RV
172
Intonation and Stress: Evidence from Hungarian
In this frame, the subscript XP represents a maximal syntactic category or a phrasal compound immediately dominating the position (marked by the dash) into which an RV word will have to be inserted. This position is preceded by a single-accented constituent (A) or a double-accented RV word (BC) within the phrase. The rule is as follows:
(34) Trochaic Reversal * 3 0 /
* B C
[Trochaic Reversal Frame] * * * * * * e.g. tizenhárom 3 tizenhárom; ugyanaz 3 ugyanaz Rule (34) creates the end-accented versions of the RV words presyntactically and ensures that these versions should be given phonological instantiation postsyntactically. Trochaic Confirmation on the other hand will be treated as a P1 (postlexical) rule:
(35) Trochaic Confirmation * 3 0 / XP[PC[* ]PC * ... ]XP Condition: XP is a maximal syntactic category starting with a double-accented phrasal compound (PC), RV or non-RV. As has been shown, precompiled rules and P1 rules do refer to lexical and syntactic information, but this is a legitimate possibility for exactly these types of rules even in models which deny the accessibility of such information for postlexical phonological rules. We shall now demonstrate how the rules work in concrete examples. In (36) the only change that can affect the double-accented RV word tizenhárom is Trochaic Confirmation. PCSR stands for “Phrasal Compound Stress Rule”. This ensures that the stressing of phrasal compounds is like that of phrases: each lexical word within the compound receives an accent.
(36) tizenhárom utas ‘13 passengers’ 10-on-three passenger PCSR: * * [tizenhárom] [B C ]
Rhythmical Variation in Phrasal Compounds 173
Trochaic Confirmation: *
*
XP[PC[tizenhárom]PC
[ [B 3
C
]
* utas]XP 3 D ]
*
XP[PC[tizenhárom]PC
[ [B
C
]
* utas]XP D ]
In the next example, (37), the double-accented phrasal compound is non-RV: ütött-kopott ‘battered’, and it also undergoes Trochaic Confirmation.
(37) ütött-kopott villamosok battered trams PCSR: * * [ütött-kopott] [B C ] Trochaic Confirmation: *
*
[ [B
C
‘battered trams’
*
3
XP[PC[ütött-kopott]PC villamosok]XP
3
]
D
*
] *
XP[PC[ütött-kopott]PC villamosok]XP
[ [B
C
]
D
]
In (38) Trochaic Reversal in the precompiled (lexical) phonology produces the end-accented version of tizenhárom and this form will be instantiated.
(38) mind a tizenhárom ‘all the 13’ all the 13 PCSR: as in (36) Frame of tizenhárom: Trochaic Reversal Frame Trochaic Reversal: * * * [tizenhárom] 3 [tizenhárom] [B Instantiation:
C
]
[B
C
* * [mind a [tizenhárom] ] [ A
[B
C
]]
]
174
Intonation and Stress: Evidence from Hungarian
In (39) the word tizenhárom undergoes Trochaic Confirmation in its own, lower-ranking phrase (tizenhárom utasa), which meets the structural description of Trochaic Confirmation. But the higher-ranking phrase (a hajó tizenhárom utasa) does not match the structural description of Trochaic Confirmation and so Trochaic Confirmation does not take place for the second time.
(39)
a hajó tizenhárom utasa the ship 13 passenger-its ‘the ship’s 13 pasengers’ PCSR: as in (36) Trochaic Confirmation: * * * * hajó XP[PC[tizenhárom]PC utasa]XP]XP 3 [ A [ [B C ] D ] ]
XP[a
3
* * * hajó XP[PC[tizenhárom]PC utasa]XP]XP [ A [ [ B C ] D ] ]
XP[a
In (40) the word tizenhárom goes through Trochaic Reversal after száz‘hundred’ and thus the word száztizenhárom becomes derivatively double-accented: [száztizen[három. This is reanalysed as [B C] and submitted to Trochaic Confirmation in the phrase száztizenhárom utas.
(40)
száztizenhárom utas ‘113 passengers’ 100 13 passenger PCSR: as in (36) Frame for tizenhárom: Trochaic Reversal Frame Trochaic Reversal: * * * [tizenhárom] 3 [tizenhárom] [B C ] [B C ] Instantiation: * * [száz[tizenhárom]] [A [B C ]] Reanalysis: * * * * [száz[tizenhárom]] 3 [száztizenhárom] [A [B C ]] [ B C ]
Rhythmical Variation in Phrasal Compounds 175
Trochaic Confirmation: * * * XP[PC[száztizenhárom]PC utas]XP 3 [ [ B C ] D ] * * 3 XP[PC[száztizenhárom]PC utas]XP [ [ B C ] D ] In the next example, (41), the word tizenhárom undergoes Trochaic Confirmation, while the word negyvenöt preserves both its accents. The output of (41) cannot be submitted to Trochaic Reversal because that would mean going back from the postlexical (P1) stratum to lexical (precompiled) phonology, which cannot be done.
(41)
tizenhárom negyvenöt ‘13:45’ (time indication) 13 45 PCSR: * * * * [tizenhárom], [negyvenöt] [ B1 C1 ] [ B2 C2] Trochaic Confirmation: *
*
XP[PC[tizenhárom]PC
[ [ B1 C1 3
]
* * [negyvenöt]]XP 3 [ B2 C2 ]]
*
XP[PC[tizenhárom]PC
[ [ B1
C1
]
* * [negyvenöt]]XP [ B2 C2]]
As an alternative to (41), PCSR can be first followed by Trochaic Reversal on negyvenöt, and then by Trochaic Confirmation on tizenhárom, as is shown in (42). Since the rules here affect concatenated RV words, there is no restriction on the number of inter-accentual syllables (section 7.2.2).
(42)
tizenhárom negyvenöt PCSR: * * * * [tizenhárom], [negyvenöt] [ B1 C1 ] [ B2 C2] Frame for negyvenöt: Trochaic Reversal Frame Trochaic Reversal:
* * * [negyvenöt] 3 [negyvenöt] [ B2 C2 ] [ B2 C2 ]
176
Intonation and Stress: Evidence from Hungarian
Instantiation:
* * * [[tizenhárom] [negyvenöt]] [[ B1 C1 ] [ B2 C2 ]]
Trochaic Confirmation: * * [[tizenhárom] [[ B1 C1 ]
* [negyvenöt]] 3 [ B2 C2]]
* 3 [[tizenhárom] [[B1 C1 ]
* [negyvenöt]] 12 [ B2 C2]]
Finally, (43) is meant to show that the postlexical rule of Trochaic Confirmation operates cyclically. The generally assumed non-cyclicity of postlexical rules can only mean that the Strict Cycle Condition does not hold postlexically, and “not that the rule does not respect constituent structure” (Gussenhoven 1991: 17).
(43) ugyanaz a tizenhárom utas same the 13 passsenger ‘the same 13 passengers’ PCSR: * * * * [ugyanaz], [tizenhárom] [B1 C1] [B2 C2 ] Trochaic Confirmation-1: *
*
*
XP [ugyanaz
[ B1
a
*
XP[PC[tizenhárom]PC
C1
[ [B2
C2
]
* utas]XP]XP 3 D ] ]
* * * * 3 XP[ugyanaz a XP[PC[tizenhárom]PC utas]XP]XP [ B1 C1 [ [ B2 C2 ] D ] ] Trochaic Confirmation-2: *
*
XP[PC[ugyanaz]PC
[ [B1
C1]
* * a XP[tizenhárom utas]XP]XP 3 [ B2 C2 D ] ]
* * * 3 XP[PC[ugyanaz]PC a XP[tizenhárom utas]XP]XP [ [B1 C1] [ B2 C2 D ] ]
Rhythmical Variation in Phrasal Compounds 177
7.4
Summary
This chapter has dealt with the two kinds of rhythmical variation observable in the stressing of Hungarian phrasal compounds which are embedded in larger phrases. The chapter has demonstrated that models using a fully relational stress system, which operate with relative intensity differences rather than with the presence or absence of pitch accent, fail to account for the facts of Hungarian rhythmical variation. However, models like the one proposed in this chapter, which disregard the relative strengths of major stressed syllables and emphasize the absolute (“either–or”) nature of major stresses, viz. the fact that they are pitchaccents, can do this job adequately. The analysis proposed preserves the advantages of the accent-based approach developed in Gussenhoven (1991), but it splits the unitary account offered by Gussenhoven (for English) and offers two rules for Hungarian. The Split Analysis expresses the asymmetry that exists between the two kinds of rhythmical variation in Hungarian. The most important conclusion of this discussion is that the phonetic intensity differences between major stresses are irrelevant in accounting for Hungarian phonological operations like rhythmical variation.
8 Rhythmical Secondary Stresses
8.1
Introduction
In the examples of Chapters 6 and 7, minor stresses were derived by reduction (deaccentuation) from major stresses that appeared on the first syllable of words. Now we shall move on to another kind of minor stress, which is not derived from major stress but is assigned to certain unstressed syllables postlexically for rhythmical reasons. This is the kind of stress that, in some renderings at least, can be identified on the third, fifth and seventh syllables of the word amerikaiakat when it is pronounced in isolation, see (1). (The numerals below the word show the number of syllables.)
(1) [ame]rika]ia]kat 1
‘Americans-ACC’
2 3 4 56 7
Since these stresses can apparently occur in isolated words, they are traditionally referred to as secondary stresses in the literature, and we are not going to deviate from this usage, although strictly speaking, as we shall see, they are not word-level minor stresses. We shall call them rhythmical secondary stresses in the rest of this chapter. It has been traditionally assumed that Hungarian non-compound words have rhythmical secondary stresses on certain syllables after the initial syllable, which is primary-stressed; see, for example, Hall (1944: 17), Papp (1966: 159). Such secondary stresses occur on syllables that are not adjacent to primary- or secondary-stressed syllables. Such views may create the impression that these secondary stresses are an inherent property of Hungarian words, and are just as obvious, permanent and 178
Rhythmical Secondary Stresses 179
automatic features of the words in Hungarian as they are in, for example, Maranungku.1 This assumption will be refuted in this chapter. After reviewing the relevant literature in section 8.2, in section 8.3 we will show that there is a rule, Contour Insertion, that seeks minorstresses. Contour Insertion will be used as a diagnostic device to help find minor stresses (including rhythmical secondary stresses). Then, in section 8.4, it will be shown that the domain of rhythmical secondary stress placement is not the word but the intonation phrase, and rules will be proposed for the two patterns of rhythmical secondary stress assignment in Hungarian.
8.2
Earlier accounts
Hayes (1995: 330) finds that the literature shows two patterns of rhythmical secondary stress placement in Hungarian non-compound words. In both patterns the primary stress falls on the first syllable. In Pattern A, originally described by Balassa (1890), and made available for English speaking scholars by Kerek (1971), the secondary stresses follow the primary stress on every odd-numbered syllable. In the other pattern, Pattern B, described by Szinnyei (1912: 12), “a secondary stress falls on the third and the fifth syllables or (if the third syllable is light) the fourth and sixth, but never on the last” (translation from Hayes, 1995).2 Kager (1995: 374), and Roca and Johnson (1999: 347) do not recognize Pattern B and speak of Pattern A as the only possibility. Hammond (1987) also assumes that this is the only pattern, but he differentiates two degrees of secondary stresses. He achieves this by superimposing upon the pattern a special colon layer which assigns greater prominence to odd-numbered feet, as, for example, in [féle]mele; teid ‘your mezzanins’ (where I use the semicolon to indicate the stronger secondary stress). But this differentiation of secondary stresses into stronger and weaker degrees has little justification. Native Hungarian speakers simply do not hear the alleged differences and find Hammond’s examples unconvincing. While Hammond overdifferentiates rhythmical secondary stresses, Kálmán and Nádasdy (1994: 407) deny their existence altogether. Further uncertainty comes from the modifications that have been proposed to the stress patterns described, on the basis of some, rather vaguely established, morphological criteria. For instance, Szende (1976: 120) claims that so-called “living” (i.e. productive) derivational suffixes may attract secondary stress, even if they are not rhythmically eligible (odd-numbered) syllables; as, for example, the suffix -ság ‘-ness’ has
Intonation and Stress: Evidence from Hungarian
180
secondary stress in the word [hatalmas]ságok ‘mighty ones’ (literally ‘mightinesses’). Gráf (2001) establishes feet and thereby rhythmical secondary stresses in words by applying a set of constraints and assuming that such stresses are sensitive to syllable- and foot quantity. However, Gráf’s proposal has not been experimentally substantiated and so, at this stage, it is no more than a hypothesis, albeit a feasible one. In the next section I will show that certain minor stresses, including rhythmical secondary stresses, may become the initiators of a high monotone intonation contour, and thus may become major stresses. I call this phenomenon Contour Insertion. Contour Insertion can be used as a diagnostic test for minor stresses, including rhythmical secondary stresses, in Hungarian.
8.3
Contour Insertion
8.3.1
The mechanism and meaning of contour insertion
Contour Insertion can happen in a Hungarian intonational phrase if the following conditions are met:
(2) Conditions for Contour Insertion a. The penultimate major-stressed syllable within the IP initiates a half-falling contour; b. the last major-stressed syllable within the IP initiates a fully falling contour; c. there are three or more unaccented syllables between these two major-stressed syllables; d. at least one of these unaccented syllables is minor-stressed. If these criteria are met, the minor-stressed syllable (or one of the minor-stressed syllables if there is more than one) between the original two major stresses may become the starting point of a high monotone intonation contour, as is shown graphically in (3).
(3) 3
Rhythmical Secondary Stresses 181
The introduction of this high monotone between the curtailed half fall and the full fall is what I call Contour Insertion. As an example, consider (4). In the normal version, (4)n, the minor stress is non-rhythmical, it occurs on the first syllable of the personal pronoun (polysyllabic function word) nektek. This is the place for Contour Insertion, shown in (4)a.
(4) Elmondok
nektek
PERF-tell-1sg you-pl-to ‘I’ll tell you a story.’
egy
történetet.
a
story-ACC
The normal version: n. | [Elmondok ]nektek egy `történetet. | El tör mondok nektek egy téne tet. The contour-inserted version: a. | [Elmondok 4nektek egy | `történetet. | El
nektek egy mondok
tör téne tet.
By initiating the high monotone, the minor-stressed syllable automatically becomes major-stressed, because now it initiates a character contour in addition to having extra intensity.3 Contour Insertion divides the original IP into two derived IPs because the high monotone is a terminal contour (see section 3.6), and so its presence implies the presence of an IP boundary. A derived IP does not have to satisfy Selkirk’s (1984: 290–6) “Sense Unit Condition on Intonational Phrasing”. Using Bolinger’s (1986, 1989) terms, Contour Insertion changes a sequence of A + A profiles into a sequence of A + B + A profiles over the same string of syllables. The result is the same as that of Replacement of Half Fall by High Monotone (see example (61) in section 5.4.7), and the attitude conveyed by it is also the same: “schoolmasterly”. The sequence of half fall, high monotone, full fall at the end of an utterance so obtained signals that the speaker considers the contents of the sentence self-evident, usually with an implication of the speaker’s intellectual superiority or official authority and it sounds patronising and categoric but not unfriendly. For instance, (4)a is possible in a school context, (5), where the teacher is asking the children what they think she is going to do next, and is answering her own question:
182
Intonation and Stress: Evidence from Hungarian
(5) (Na és mit gondoltok, most mit csinálok?) | [Elmondok
4nektek
egy
| `történetet. |
‘(And what do you think I’ll do next?) I’ll tell you a story.’ If the domain contains more than one non-rhythmical minor stress, it seems that it is the last one of these that receives Contour Insertion. In (6) there are two words starting with a non-rhythmical minor stress: volna and magát. The last one will receive Contour Insertion:
(6) Nem not
tudta
volna
magát
megmenteni.
could-3sg
would-be
himself
save-INF
‘He couldn’t have saved himself.’ The normal version: n. | [Nem tudta ]volna ]magát `megmenteni. | The contour-inserted version: a. | [Nem tudta ]volna 4magát | `megmenteni. | b. *?| [Nem tudta 4volna ]magát | `megmenteni. | In (4) and (6) the minor stress which Contour Insertion has found is non-rhythmical. If there is no non-rhythmical minor stress in evidence, but Contour Insertion is still possible, as in (7), we must assume that it targets a rhythmical secondary stress. The numerals below the examples show the number of syllables within the penultimate contour carrier.
(7) Elmegyek
a
away-go-1sg
könyvtárba.
the library-to
‘I am going to the library.’ The normal version: n. | [Elmegyek a `könyvtárba. | 1
2
3
4
The contour-inserted version: a. | [Elme4gyek a | `könyvtárba. | 1
2
3
4
In (7)n there were only three unaccented syllables between the original two major stresses, the penultimate contour carrier was four syllables
Rhythmical Secondary Stresses 183
long. As we have seen, in this case the third syllable of this domain can receive Contour Insertion: (7)a. No other syllable is eligible for it; neither (7)b nor (7)c is acceptable.
(7) b. * | [El4megyek a | `könyvtárba. | c. * | [Elmegyek 4a | `könyvtárba. |
Since the third syllable after the primary stress is associated with rhythmical secondary stress (see section 8.2), I assume that the third syllable in (7)n may receive Contour Insertion because it is secondary-stressed. However, in cases where the original domain of the contour is more than four syllables long, the place of Contour Insertion is less obvious than in cases such as (7). We may continue to assume that it is a rhythmical secondary stress that attracts Contour Insertion in these cases, too. In order to check the validity of this assumption, I have carried out an experiment (Varga 2000). 8.3.2
The experiment
In the experiment I tape-recorded my own renderings of four blocks of sentences. Each block contained a certain number of identical sentences with different placements of Contour Insertion in them. Each rendering of each sentence was made in such a way that no stress clashes arose, and so none of the sentences was totally unacceptable. 4 The blocks were played to 105 university and college students, all native speakers of Hungarian, aged between 18 and 36. The renderings in the blocks were numbered. The participants were given questionnaires with the numbers of the renderings on them and, after hearing each rendering twice, they were asked to write against each entry whether or not they had found it acceptable. Then, after listening for a third time, the participants were asked to identify the rendering which they thought was the best in its block. Each acceptance was worth 1 point, and each “best” vote was also worth 1 point. All the points given to a rendering were totalled and divided by 105 to obtain the mean score for that rendering. Finally, the mean scores of the various renderings were submitted to statistical analysis in order to see whether they differed significantly. T-test statistics were applied in the case of two variants, and one-way Analysis of Variance (ANOVA) was used for multiple variants. 5 The experiment has provided evidence that, in cases where the original domain of the contour is more than four syllables long, the preferred
184
Intonation and Stress: Evidence from Hungarian
place of Contour Insertion is still one of the rhythmically secondarystressed syllables within the domain. In the following I will show the sentences used in the experiment. The renderings of each sentence will be shown in the order of decreasing score values. In (8) the original domain of the half fall was five syllables long. The mean scores differed significantly at the p < .05 level, which means that (8)a, with a mean score of 0.83, was significantly more acceptable to the listeners than (8)b.6
(8) A the
barátaimnak
Jóska.
friends-my-DAT
Joe
‘To my friends, I am just Joe.’ The normal version: n. | A [barátaimnak `Jóska. | 1 2 34
5
The contour-inserted versions:v
Mean:
a. | A [bará4taimnak | `Jóska. | 1 2 34
0.83
5
b. * | A [baráta4imnak | `Jóska. | 12 3 4
0.40
5
In (9) the original domain of the half fall is also 5 syllables long, and it is the third syllable once again which attracts Contour Insertion with the significantly greatest acceptance ratio.
(9) A the
barátaival
nem
beszél
így.
friends-his-with
not
speaks
so
‘To his friends, he does not speak like this.’ The normal version: n. | A [barátaival `nem beszél így. | 1 2 34 5 The contour-inserted versions: a. | A [bará4taival | `nem beszél | így. |
Mean: 1.10
1 2 34 5 b. * | A [baráta4ival | `nem beszél így. | 123 45
0.60
Rhythmical Secondary Stresses 185
The results of (8) and (9) show that, when the domain of the half fall is five syllables long, it is the third syllable that will provide the significantly most acceptable contour-inserted version. The third syllable is a traditionally recognized site of rhythmical secondary stress. In the next block, (10), the number of syllables in the original domain is six.
(10) Ez
egy
this a(n)
kiapadhatatlan
eroforrás.
inexhaustible
strength-source
‘This is an inexhaustible source.’ The normal version: n. | Ez egy [kiapadhatatlan `eroforrás. | 12 3
45 6 Mean:
The contour-inserted versions: a. | Ez egy [kia4padhatatlan | `eroforrás. | 12 3
1.29
45 6
b. * | Ez egy [kiapadha4tatlan | `eroforrás. | 12 3
4 5 6
c. * | Ez egy [kiapad4hatatlan | `eroforrás. | 12 3
0.53
45 6
d. * | Ez egy [kiapadhatat4lan| `eroforrás. | 12 3
0.61
45
0.44
6
Here the ANOVA analysis revealed that (10)a differed significantly from all the other contour-inserted renderings. This means that in (10), too, it is the third syllable of the original domain that is the likeliest candidate for Contour Insertion. These results provide further evidence that rhythmically secondary stressed syllables may attract Contour Insertion. In the last block of sentences, (11), the domain of the original contour contains eight syllables.
(11) Kísérletezhetnétek experiment-could-2pl
egy
másikkal.
an
other-with
‘You could experiment with another one.’ The normal version: n. | [Kísérletezhetnétek egy `másikkal. | 1 2 34
5 6 7
8
186
Intonation and Stress: Evidence from Hungarian
The contour-inserted versions:
Mean:
a. | [Kísérletez4hetnétek egy | `másikkal. | 12 34
5 67
b. ? | [Kísérletezhet4nétek egy | `másikkal. | 12 3 4 5
67
3 4 5 67 5
67
5 6 7
0.50
8
d. * | [Kísérletezhetné4tek egy | `másikkal. | 12 3 4
0.62
8
d. * | [Kísérle4tezhetnétek egy | `másikkal. | 12 3 4
0.85
8
c. * | [Kísér4letezhetnétek egy | `másikkal. | 12
1.12
8
0.34
8
In (11) the contour-inserted solution with the significantly highest acceptance ratio (a mean score of 1.12) places the new primary stress on the fifth syllable of the penultimate contour carrier (11)a, which corresponds to a traditionally recognized secondary-stressed position. (The fifth syllable wins over the third because the fifth syllable is in the middle of the original domain and cuts the latter into two equal parts.) However, a number of speakers would put the new primary stress even further to the right: on the sixth syllable of the penultimate contour carrier (11)b. This mean score (0.85) is relatively high, and differs significantly from all the other lower scores at the p < .05 level. A possible explanation may be that these speakers use Pattern B of secondary stress placement (see section 8.2). In this pattern the secondary stress may fall on the fourth and sixth syllables if the third syllable is light (which it is: -le-). I do not know why it is the sixth, rather than the fourth, syllable that attracts Contour Insertion in (11)b; for some reason a 5–3 syllable division is judged to be preferable to a 3–5 syllable division. Table 8.1 summarizes the results of the statistical analysis of the mean scores. In all the “winning” examples (8)a, (9)a, (10)a and (11)a, Contour Insertion has appeared on one of the odd-numbered syllables, that is on one of the syllables that are traditionally considered rhythmically secondary-stressed in Hungarian words. In (11)b, a “second-best” solution, the syllable chosen is an even-numbered one, but this, too, can be regarded as rhythmically secondary-stressed, in Pattern B. On the basis of the examples we have examined we can define Contour Insertion as shown in (12). However, as will be soon clear, this
Rhythmical Secondary Stresses 187 Table 8.1 Sentence
Summary of the statistical analysis of mean scores Mean
Standard
Probability
Significance
8a 8b
0.83 0.40
0.68 0.63
0.000
*
9a 9b
1.10 0.60
0.69 0.67
0.000
*
10a 10b 10c 10d
1.29 0.61 0.53 0.44
0.78 0.69 0.66 0.58
0.000
*
11a 11b 11c 11d 11e
1.12 0.85 0.62 0.50 0.34
0.75 0.67 0.78 0.60 0.56
0.000
* (*)
definition does not cover all cases of Contour Insertion, and so it should be regarded as a preliminary definition.
(12) Contour Insertion (Preliminary) ]σ . . . =opt: 4σ . . . | /[σ . . . ___ `σ ( . . . ) | where σ = syllable, ]σ = minor-stressed syllable, ... = a string of one or more syllables with no major stress or IP boundary in it. Condition:
“schoolmasterly” attitude.
Comments: (a)
If the string of syllables carrying the first contour contains non-rhythmical minor stresses, it is the last one of these that may host Contour Insertion. (b) If the string of syllables carrying the first contour contains only rhythmical secondary stresses, it is one of these that may host contour insertion. This may be the 3rd or 5th syllable (or – for a smaller number of speakers – the 4th or 6th syllable if the 3rd syllable is light), with preference for the syllable which is closer to the middle of the original half falling domain.
188
8.3.3
Intonation and Stress: Evidence from Hungarian
Extension
Up to this point we have only seen cases where Contour Insertion has taken place on a minor-stressed syllable within the domain of a contour which immediately preceded a full fall. In other words, it appears that Contour Insertion can happen only once in an utterance and only immediately before a final full fall. This is not so. Consider the examples in (13). In these examples, in addition to the normal version (13)n and the simple contour-inserted version (13)a, other contour-inserted versions (13)b and (13)c are also possible: (13)
Azt
gondolom, hogy
that=DEM-ACC think-1sg
nem kellene
that=CONJ not
veszekednünk.
ought-to quarrel-INF-1pl
‘I don’t think we should quarrel.’ (Lit.: ‘I think that we should not quarrel.’) The normal version: n.
| [Azt gon]dolom, hogy [nem kel]lene `veszekednünk. | Azt gondolom, hogy
nem
kellene vesze
kednünk.
The contour-inserted versions: a.
| [Azt gon]dolom, hogy [nem kel4lene | `veszekednünk. | Azt gondolom, hogy
b.
lene kel
ve szeked
nünk.
| [Azt gon4dolom, hogy | [nem kel]lene `veszekednünk. | Azt gon
c.
nem
dolom, hogy nem
kellene ve sze
kednünk.
| [Azt gon4dolom, hogy | [nem kel4lene | `veszekednünk. | Azt
dolom, hogy gon
nem
lene ve kel
szeked nünk.
In (13b) Contour Insertion occurs not before a final full fall but before a half fall. In (13)c Contour Insertion occurs twice. In order to be able to provide a unitary account of examples such as (13)b and (13)c, as well as (13)a and all the previous examples of Contour Insertion, we have to go back to the notion of the complementary block (CB). In Chapter 5, section 5.4.4, we defined complementary blocks as those major-stressed syntactic constituents that are the immediate or non-immediate constituents of a highest-ranking sentence, between which there is a
Rhythmical Secondary Stresses 189
complementary relationship. For instance, in (13)n we have the following CBs (nf stands for ‘non-final’, f stands for ‘final’):
(13) n. | [Azt gon]dolom, hogy [nem kel]lene `veszekednünk. | CBnf CBnf
CBf CBf
CBf = Sentence It seems, then, that Contour Insertion is possible on a minor-stressed syllable within the tail of any half falling contour that stands immediately before a final CB that has a fully falling contour. In (13) there are two such sentence-final CBs: veszekednünk and nem kellene veszekednünk, and Contour Insertion is possible before the smaller one (13)a, or before the larger one (13)b, or before both (13)c. On the basis of these facts Contour Insertion can be formulated in a more general fashion, which covers all possible cases. This is given in (14):
(14) Contour Insertion (Extended) ]σ . . . =opt: 4 σ . . . | / [σ . . . ___ [f(Y) `σ ( . . . ) | ] where σ = syllable, ]σ = minor-stressed syllable. Y = material of the final CB up to the falling contour, . . . = a string of one or more syllables with no major stress or IP boundary in it. Condition: “schoolmasterly” attitude. NB: As to which minor-stressed syllable should host Contour Insertion, see the Comments in (12). Contour Insertion is important because it can be used as a diagnostic technique for finding minor stresses (including rhythmical secondary stresses which occur between two major stresses). The fact is that secondary stressed syllables are not always obvious, if their non-melodic prominence is slight. Therefore the significance of Contour Insertion in Hungarian parallels that of Iambic Reversal or the Stylized Fall for the identification of minor stresses in English words.7 In light of this, I cannot agree with views that deny the existence of rhythmical secondary stresses and the phonological relevance of minor stresses in Hungarian, saying that no phonological rule is sensitive to the presence or absence of minor stresses (e.g. Kálmán and Nádasdy 1994: 406). Contour Insertion
190
Intonation and Stress: Evidence from Hungarian
is just such a rule: it is sensitive to minor stresses. Hungarian minor stresses are phonologically relevant. The experiment also demonstrated that rhythmical secondary stresses are a reality in Hungarian; that both Pattern A and Pattern B exist, but that Pattern A prevails.
8.4 The domain of rhythmical secondary stress placement Although rhythmical secondary stresses may appear in the isolated pronunciation of individual words, in reality they depend on foot boundaries rather than on word boundaries. It may happen that an unbounded foot contains more than one word because one of them is a clitic or has been destressed by Clash Deletion (see section 6.6.3). So rhythmical secondary stresses need not coincide with the odd-numbered syllables of words, as indeed they do not in (15).
(15) Azt that-ACC
mondtam,
hogy
elég.
said-1sg
that
enough
‘I said that it was enough.’ The normal version: n. | [Azt mond]tam, hogy `elég. | 1
2
3
4
The contour-inserted versions: a. | [Azt mond4tam, hogy | `elég. | 1
2
3
4
In (15)n the third syllable of the penultimate contour carrier coincides with the second syllable of the word mondtam, that is the secondary stress falls on an even-numbered syllable within that word. This syllable has received its secondary stress not as a syllable of a word but as a syllable of a larger-than-word-size prosodic constituent. This shows that rhythmical secondary stress placement must be a postlexical process in Hungarian; such secondary stresses are superimposed upon strings of words by reparsing the unbounded feet present in the string into leftheaded binary feet, from left to right. (On how the unbounded feet come into being in the first place, see Chapter 6, section 6.5.1.) We claim that the domain of this reparsing (i.e. the domain of rhythmical secondary stress placement) is the intonational phrase, as defined
Rhythmical Secondary Stresses 191
in sections 3.6 and 4.10. This unit has a characteristic internal structure, and can be derived by an algorithm based partly on syntactic and partly on prosodic structure (see Chapter 5). As was shown in section 3.6.3, the Hungarian IP contains a certain number of unbounded feet (plus an optional unfooted clitic in IP-initial position). We also saw that major-stressed feet, together with all the minor-stressed feet on their right, constitute accent phrases. (These are missing in IPs that consist of appended contours.) For instance, (16) is an IP which consists of an unfooted clitic, four feet (each starting with a syllable with a left parenthesis), and three accent phrases (each starting with an asterisked syllable). The unfooted clitic at the beginning constitutes the preparatory part of the IP, the first and second accent phrases form the scale, and the third accent phrase is the terminal part of the IP.
(16) De eredetiben olvasta az angol regényeket. but original-in read-3sg the English novels-ACC ‘But he read the English novels in the original.’ |De [eredetiben ]olvasta az [angol `regényeket.| * * * [De (eredetiben (olvasta az (angol (regényeket.] The intonational phrase is submitted to Rhythmical Secondary Stress Assignment (RSSA), that is rule (17). The RSSA is a foot-constructing rule, which will place rhythmical secondary stress on every second syllable from left to right starting from a primary- or secondary-stressed syllable within the intonational phrase. In its effect, the RSSA is like Halle and Vergnaud’s (1987) “Alternator”, or Halle and Idsardi’s (1995) or Halle’s (1998) “Iterative Constituent Construction”, or Hayes’ (1995) “Strong Local Parsing”, but it is a postlexical rule. Rhythmical secondary stresses in Hungarian are assigned postlexically.8
(17) Rhythmical Secondary Stress Assignment (RSSA) * 0 =obl : ( / IP[ . . . ( σ σ ___ σσ0 . . . ]
σ
Reparse each unbounded foot within an intonational phrase into left-headed binary feet, by inserting a left parenthesis after every two syllables, from left to right. Condition: a foot can only be constructed if its appearance does not create a stress clash within the intonational phrase. Domain: intonational phrase.
192
Intonation and Stress: Evidence from Hungarian
On the basis of the next example, (18), it might at first appear that the domain of RSSA is the unbounded foot (Ft), and not the IP. But even this example is compatible with the assumption that the domain of RSSA is the IP, since the Ft in it is coextensive with an IP.
(18) Sört ittam. beer-ACC drank-1sg ‘I drank beer.’ a. [Sört ittam. b. [Sört it]tam. 3 Ft:
* (Sört (ittam.
CD:
* (Sört ittam.
IP:
* [(Sört ittam.]
* RSSA: [(Sört it(tam.] However, examples like (19) make it clear that the real domain of RSSA is not the Ft but the IP. Here the RSSA cannot work in the first Ft (sört ittam) and cannot produce a secondary stress on the syllable -tam, because there is a major stress on the syllable reg-, which is in another Ft but in the same IP, and so it blocks RSSA.
(19) Sört ittam reggelig. morning-till ‘I was drinking beer till the morning.’ a. [Sört ittam [reggelig. 3 b. [Sört ittam [regge]lig. Ft:
* * (Sört (ittam (reggelig.
CD:
* * (Sört ittam (reggelig.
IP:
* * [(Sört ittam (reggelig.]
* * RSSA: [(Sört ittam (regge(lig.] As formulated in (17), the RSSA can also account for rhythmical secondary stresses in single words pronounced in isolation. A (non-compound) word pronounced in isolation realizes an unbounded foot which realizes
Rhythmical Secondary Stresses 193
an IP (and the IP, in turn, realizes an utterance). For instance, the word amerikaiakat ‘Americans-ACC’, which emerges from the lexical stratum as (20)a, may become (20)b when pronounced in isolation as an intonational phrase.
(20) a. [amerikaiakat 3 Ft: CD:
* (amerikaiakat n/a
IP:
* [(amerikaiakat]
b. [ame]rika]ia]kat
* RSSA: [(ame(rika(ia(kat] Those speakers who use Pattern B for rhythmical secondary stress assignment, may put the secondary stress on the fourth syllable in amerikaiakat because the third syllable is light. This version can be obtained by a special (Pattern B) version of the RSSA, given here as (21). In its effect, it is like Hayes’ (1995: 308) Weak Local Parsing, but it is postlexical.
(21) Rhythmical Secondary Stress Assignment, Pattern B (RSSA-B) When the second foot boundary has been placed in accordance with (17), and the first syllable of the second foot is light, incorporate that syllable into the preceding foot by moving the second foot boundary one syllable to the right, and then continue constructing feet in accordance with (17). Condition and domain: as in (17). The derivation is shown in (22):
(22) a. [amerikaiakat 3 b. [ameri]kai]akat Ft: CD:
* (amerikaiakat n/a
IP:
* [(amerikaiakat]
RSSA:
* [(ame(rikaiakat]
* RSSR-B: [(ameri(kai(akat]
194
Intonation and Stress: Evidence from Hungarian
When the compound [pót]kerékcsapágy has undergone Clash Deletion and become [pótkerékcsapágy (see (14)a in Chapter 6, reproduced here as (23)a), it may undergo Rhythmical Secondary Stress Assignment and become (23)b:
(23) a. [pótkerékcsapágy 3 b. [pótke]rékcsa]págy Ft:
* (pót(kerékcsapágy
CD:
* (pótkerékcsapágy
IP:
* [(pótkerékcsapágy]
* RSSA: [(pótke(rékcsa(págy] The derivation shows that CD is applied earlier than the RSSA.
8.5
The order of rules
The rules described in Chapters 6, 7 and 8 call for the following ordering.
(24) (i) Lexical Rules: a. Word Stress Rule [(19) in Chapter 6] b. Lexical Foot Erasing [(12) in Chapter 6] c. Compound Stress Rule [(10) in Chapter 6] d. Trochaic Reversal [(34) in Chapter 7] (ii) Postlexical Rules: e. Trochaic Confirmation [(35) in Chapter 7] f. Syntactic Deaccentuation [(32) in Chapter 6] g. Semantic/Pragmatic Deaccentuation [(38) in Chapter 6] h. Clash Deletion [(39) in Chapter 6] i. Rhythmical Secondary Stress Assignment [(17) in Chapter 8], with its Pattern B variant: [(21) in Chapter 8] j. Contour Insertion [(14) in Chapter 8] k. Ultimate Foot Erasing [(15) in Chapter 6]
8.6
Summary
This chapter has dealt with rhythmical secondary stresses. These are, contrary to common belief, a property of intonational phrases and not
Rhythmical Secondary Stresses 195
of words. Rhythmical secondary stresses may also appear in isolated words, creating the false impression that they are a property of words, but even in these cases they appear in IPs because words pronounced in isolation are, in fact, IPs. In this chapter we paid special attention to the status of minor stresses. Since the phenomenon of Contour Insertion, presented here, requires a minor-stressed syllable, Hungarian minor stresses are not only phonetic features of speech but are also phonologically significant. Besides, by their ability to host Contour Insertion and thus to initiate a high monotone intonational contour, certain minor stresses can be relevant to intonation.
9 Summary and Conclusions
In this chapter we shall briefly recapitulate the main points of the book, once again drawing attention to the theoretical issues addressed and to the similarities and differences between the intonational and stress systems of Hungarian and English. In Chapter 1 we advanced the view of intonation which was further elaborated in later chapters. We considered intonation to be partly determined by the surface syntax of the sentence and thus to emerge in several steps in the course of the syntax–phonology mapping. This mapping was supposed to take place in two stages: stress fixing and melodic segmentation. Chapter 2 gave a general overview of the topic of intonation. We clarified the basic concepts that were necessary for defining intonation, and accommodated intonation within the broader framework of prosodic devices. We then defined intonation as those pitch variations that resulted in discrete, categorically different vocal effects called melodic prosodemes and upstep prosodemes, functioning in utterances, and being closely related to stress prosodemes and pause prosodemes. We saw that while the prosodemes formed linguistic systems, their use in utterances was accompanied by paralinguistic prosodic devices and was constrained by non-linguistic prosodic limits. Finally, the main functions of intonation were presented and discussed. We found that Hungarian, in the same way as English, is an intonation (non-tone) language in which the linguistic use of pitch variation is restricted to the differentiation of utterances rather than words. With regard to the prosodic limits, we found that Hungarian uses a considerably lower average pitch register and a narrower range than English. As for the functions of intonation, it was pointed out that Hungarian intonation is more grammaticized than English because, for instance, it is the most important single 196
Summary and Conclusions
197
means of distinguishing yes–no questions from statements. In this respect Hungarian intonation differs significantly from English. In Chapter 3 we provided a detailed taxonomic analysis of the intonation contours (melodic prosodemes) of Standard Hungarian, based on surface contrasts. Twelve meaningful intonation contours (eleven “character contours” and one “appended contour”) were distinguished and their forms and meanings described. The result of the analysis was an “intonational lexicon” of Hungarian, in which each entry consists of a representation of the phonological shape of the contour, matched by a representation of its meaning. The chapter also dealt with the stress, pause and upstep prosodemes of Hungarian, and presented the internal structure of the Hungarian intonation(al) phrase. The so-called front-falling contours (the full fall, the half fall and the fall–rise) of Hungarian have a great degree of formal and functional similarity to their English counterparts, though the fall–rise is not normally used as the melody of “given” information (shared knowledge) in Hungarian, whereas it can be so used in English. The Hungarian sustained contours (the rise, the high monotone and the descent) are again remarkably similar to their English counterparts in both form and function, but what we called the second-type descent of Hungarian, used for evaluative exclamations, cannot be used in this way in English. The end-falling contours (the rise–fall, the monotone-fall and the descent-fall), which mark yes–no interrogative sentences in Hungarian, do not exist in English and so their transfer to English by Hungarian speakers of English is a potential source of problems. The Hungarian stylized fall is used with the same functions as its English equivalent, and there is a close formal resemblance between them, though in the English version the lower terrace starts on the post-tonic secondary stressed syllable if there is one, while in Hungarian the lower terrace is always on the last syllable of the carrier phrase. The intonational phrase of Hungarian has an internal structure consisting of an optional preparatory part, optional scale and obligatory terminal part, resembling the structure of the English tone-group (optional prehead, optional head, obligatory nuclear part). However, the syllable which starts the last contour (the contour in the terminal part) in the Hungarian IP is only one of the equally prominent accented syllables within the IP and not the most prominent one. We identify the terminal contour of a Hungarian IP not on the basis of its prominence but on the basis of its shape. This is a significant difference between the IPs of Hungarian and of English. Chapter 4 was devoted to a deeper, autosegmental reanalysis of Hungarian intonation; it revisited the system of Hungarian intonation
198
Intonation and Stress: Evidence from Hungarian
contours and analysed their genesis and form, and the structure of the Hungarian intonational phrase, in an autosegmental framework established along the lines suggested by Gussenhoven (1983a, 1985). The chapter illustrated a way in which the twelve intonation contours can be derived from three basic contours just as complex words are derived from simple words by the addition of derivational affixes. It also demonstrated how the autosegmental representations of the intonation contours can be associated with the syllables of the carrier phrases and pointed out some variations in the phonetic realizations of the contour representations as well. For the representation of Hungarian basic contours the Obligatory Contour Principle had to be suspended: the H*.H.L$ (monotone-fall) and H*.L$ (half fall) represent different contours. The single representation H*.L$ for both contours would obey the OCP but would be inadequate in monosyllabic contour carriers, where the tone-syllable association cannot be manipulated to produce melodic differences. In the autosegmental framework presented in Chapter 4, no independently variable phrase accent and boundary tone were separated within the terminal contours – as they would be in Pierrehumbert’s (1980) framework – because recognizing them would have led to problems in the Hungarian end-falling contours. (The Pierrehumbertian analysis of the Hungarian rise–fall, L*HL%, with H being a phrase accent and the final L a boundary tone, is contradictory, since the final L% boundary tone, after an H phrase accent indicates a level sustension of the previous H tone, whereas we have a definite fall in the Hungarian rise–fall.) We also found that (some version of) the traditional notion of the syllable has to be retained for intonational and stress statements to be made adequately, and that recent theories which replace syllable strings with CV strings, however advantageous they may be in other areas, are simply unsuitable in their present form for investigating or describing intonation and stress. Chapter 5 examined the relation of intonation to syntax and attempted to establish rules for the “melodic segmentation” of the Hungarian utterance. Although this stage comes after the stage of “stress fixing” in the syntax–phonology mapping, the economy of description forced us to deal with melodic segmentation first and postpone the discussion of stress fixing until Chapter 6, section 6.6. The stage of melodic segmentation is responsible for dividing the utterance into intonational phrases (IPs). The IPs were approached “from above” rather than “from below”: instead of building them up from hierarchically organized lower prosodic units, they were produced by dividing the
Summary and Conclusions
199
highest-ranking prosodic unit, the utterance, into its immediate prosodic constituents, the IPs, according to certain principles. The chapter first dealt with the major tonosyntactic blocks: the melodically autonomous blocks within utterances, such as highest-ranking sentences and inorganic blocks. This was followed by a detailed analysis of various types of minor tonosyntactic blocks (i.e. major-stressed syntactic constituents within the autonomous blocks), whose common characteristic is that each type occurs in a series of sister blocks, of which the final block melodically constrains its non-final sister block(s), that is there is a right-toleft melodic dependency between the non-final and final blocks in each series. This is a major claim of the present book: that the intonation contours within the spoken realization of a sentence depend on each other in certain ways, and these dependencies can be expressed in rules. The establishment of these rules was the central concern of Chapter 5. It was established that the ultimate domain of Hungarian speech melody is the autonomous block, such as the highest-ranking sentence. It has often been claimed that – in English – any contour can be used over any stretch of the utterance as long as it expresses the speaker’s attitude or the speaker’s assessment of the stretch as being new or given information for the hearer. This principle of “any contour anywhere” was criticized by Bing (1980), who claimed that the ultimate domain of intonation is the highest-ranking sentence. This is certainly the case in Hungarian. It remains for the future to examine to what extent melodic dependencies exist within the autonomous blocks of other languages, and whether they operate from right to left, or left to right, or in a mixed fashion. In any case, the notions of “tonal sequences” in Cruttenden (1997: 103–4), or “tone concord” in Halliday (1967: 35) suggest that some kinds of melodic dependency do exist in English, though not necessarily from right to left. It also has to be mentioned that the regularities of melodic dependency may be concealed by the accidental features of performance and so for this kind of investigation the loud reading of written sentences can prove to be the best method of eliciting data. Chapter 6 started by elaborating on the three phonological degrees of stress, first introduced in Chapter 3. A major-stressed syllable is distinguished by its extra intensity and the fact that it initiates a so-called character contour (it is “pitch-accented”). A minor-stressed syllable has extra intensity but does not initiate a character contour, and a zerostressed syllable has neither extra intensity nor character contour
200
Intonation and Stress: Evidence from Hungarian
initiation. Hungarian stress degrees are not fully relational. They are relational inasmuch as major- and minor-stressed syllables have more intensity than the zero-stressed syllables, and they are absolute inasmuch as major-stressed syllables initiate character contours and so carry a pitch accent while the other syllables do not. Hungarian words, apart from phrasal compounds, have one major (primary) stress, which is on the first syllable of the word. Consequently the major stress in words is demarcative rather than distinctive in Hungarian. It cannot distinguish the meaning or syntactic class of a word as it can in English (consider [importN vs. im[portV). On the other hand, it can distinguish a word from a combination of two words built up of the same segmental phonemes, the first of which is zero-stressed: [halát ‘his fish-ACC’ from ha [lát ‘if he can see’, [akar ‘he wants to’ from a [kar ‘the arm’. One of the chapter’s main claims was that lexical words and polysyllabic function words leave the lexical stratum of phonology with a major stress, and when they are inserted in phrases and sentences, they may keep this major stress or have it reduced to minor stress, depending on whether or not the various deaccentuation processes that are due to take place in the stage of “stress fixing” are relevant to them. In the generative literature, function words are usually treated as emerging from the lexicon with zero stress. This was not accepted for Hungarian because polysyllabic function words are regularly minor-stressed on their initial syllable. We followed Selkirk (1984) in assuming that function words, too, leave the lexical phonology with major stress but are then deaccented. In the second half of Chapter 6 we examined the stressing of Hungarian phrases and sentences, that is the (largely deaccentuational) processes that take place in the stage of stress fixing within the syntax–phonology mapping. Hungarian phrases and sentences can have several, roughly equal major stresses. The chapter also discussed some of the types of prosodic contituents that we found relevant in accounting for stress and intonation. These are the syllable, the foot, the accent phrase, the intonational phrase (which we defined differently from Vogel 1987), and the utterance. Chapter 7 presented evidence for the claim that it is the pitchaccented nature of Hungarian major stresses that matters phonologically and not their intensity differences. The evidence comes from the rhythmical variation of phrasal compounds like [tizen[három ‘13’ in Hungarian. The processes of this rhythmical variation, Trochaic Reversal (e.g. [pont tizen[három = ‘exactly 13’), and Trochaic Confirmation (e.g. [tizenhárom [pont = ‘13 points’) can only be treated adequately if the non-relative side of major stresses (the fact that they initiate intonation
Summary and Conclusions
201
contours) is taken into consideration. Any attempt to account for Hungarian rhythmical variation in terms of the relative prominence (intensity) differences between major stresses inevitably breaks down. The chapter showed that rhythmical variation in Hungarian differs from its counterpart in English in two respects. First, Hungarian rhythmical variation is lexically constrained, whereas the parallel processes in English (Iambic Reversal and Iambic Confirmation) are not restricted to certain classes of words but are extremely general processes (though not without a few lexical exceptions). Secondly, the two processes of Hungarian rhythmical variation are asymmetrical: while Trochaic Reversal is restricted to rhythmically variable phrasal compounds, Trochaic Confirmation can affect all phrasal compounds, of which the rhythmically variable ones are a subset. Such asymmetry does not exist in English; the same set of words is targeted by both Iambic Reversal and Iambic Confirmation. Therefore Hungarian rhythmical variation had to be split into a lexical (precompiled) component (in the sense of Hayes 1990) producing Trochaic Reversal, and a postlexical (P1) component (in the sense of Kaisse 1985, 1990) producing Trochaic Confirmation, while English rhythmical variation could be treated as a single postlexical operation in both of its versions. At a more general level, Chapter 7 justified the claim that major stresses in Hungarian are not to be phonologically distinguished according to their intensities. This has an important consequence: it eliminates the necessity of using metrical grids to represent Hungarian metrical structure. Again, an examination of a large number of other languages would be necessary to see whether or not such behaviour occurs elsewhere, but this kind of survey would be beyond the scope of the present book. It seems likely, though, that Hungarian is not alone in this respect: intuitively the major stresses of Finnish, a language related to Hungarian, should be regarded as phonologically equal. If this proves to be correct, then the metrical grid as an analytical tool can be dispensed with in Finnish phonology, too. Chapter 7 also revealed that Hungarian cannot be easily accommodated in the syllable-timed vs. stress-timed dichotomy. It is certainly not stresstimed but, given its long and short vowels, it is not purely syllabletimed, either, even though it shows a relative constancy in the duration of syllables and a lack of systematic vowel reduction in unstressed position. Chapter 8 examined the status of rhythmical secondary stresses in Hungarian. It was claimed that rhythmical secondary stresses, which are often treated in the literature as obligatory features of the Hungarian
202
Intonation and Stress: Evidence from Hungarian
word, are the result of a postlexical process and a property of intonational phrases rather than of words. The chapter then turned to a discussion of Contour Insertion, which, in its simplest form, means assigning a high monotone contour to a minor-stressed syllable which is in the tail of a half fall that occurs before a full fall in an intonation phrase. When the originally minor-stressed syllable receives this high monotone contour, it automatically becomes major-stressed, and the original single intonation phrase is divided into two (because the high monotone is a terminal contour and ends with an IP boundary). The result of Contour Insertion is a special attitudinal variant, labelled “schoolmasterly”. Contour Insertion, being sensitive to minor stresses, provides evidence for the phonological significance of such stresses, which has been questioned recently (by Kálmán and Nádasdy 1994). It can, therefore, be used as a diagnostic technique for minor stress in Hungarian, just as Iambic Reversal and the lower terrace of the Stylized Fall can be used as diagnostic tests for secondary stress in English.
Notes
Notes to Chapter 1: Introduction 1. See Gráf (2001) on Hungarian word stress. 2. But see Hunyadi (1999) for a different view, according to which LF (logical form) influences prosody directly. 3. Accents on certain vowel letters in Hungarian orthography (as in, for example, léphetnek or híd) indicate length of the vowel, not stress or melody. (See Hungarian Letter-to-Sound Correspondences on p. xvi.) 4. Syntactic structures like this will be explained in section 1.2.3. 5. More precisely, this kind of speech can be described as belonging to Educated Colloquial Hungarian, “the spoken language of educated people in Budapest”, the capital of Hungary (Nádasdy 1985: 225). However, from an intonational point of view this further specification seems unnecessary. 6. See Kontra (1988). For a review of this project, see Nádasdy (1995). 7. This translation is from F. Juhász (1968).
Notes to Chapter 2: Intonation, Paralanguage, Prosody 1. Pitch, loudness and quality are perceptual categories, whose acoustic counterparts are fundamental frequency (Fo), intensity and the spectrum, respectively. Duration is used in both perceptual and acoustic senses. 2. Although Cruttenden (1997: 173) does not regard the pause as a prosodic device, others (e.g. Crystal 1969: 128) do. The pause as a prosodic device can be defined as that section of the speech signal which has zero intensity (Bolla 1980). 3. In tone languages pitch patterns are also used for distinguishing words: they constitute lexical tones. Hungarian has no lexical tones, only intonation: it is not a tone language but an intonation language. 4. Paralanguage can be defined in different ways. For instance, it can be defined as that part of non-segmental phonation which is beyond prosodemes (see, e.g., Gumperz and Kaltman 1980), but also as the part of non-segmental phonation which is beyond prosodic devices, if such a distinction is made (e.g. Cruttenden 1997: 174–5). In the latter case paralinguistic features are interruptive: they include vocalizations, that is meaningful but non-verbal sound combinations such as, for example [pϕ] expressing contempt; whereas prosodic features are superimposed on sound strings and occur simultaneously with them. I am in favour of the Gumperz–Kaltman view and adopt Ladd’s criterion: paralanguage has a scalar or gradient nature while linguistic messages have a quantal or categorical structure (Ladd 1996: 36). 5. We are still a long way away from a precise identification of attitudes. The labels used to refer to them are too numerous, they often overlap and their semantic reliability is questionable (Crystal 1969: 294–7). At one time it was 203
204
6.
7.
8.
9.
Notes hoped that more objective results would be obtainable from attitude scale investigations (Uldall 1960, 1964), but these were not carried out on a large scale. While the traditional classification of sentences according to modal type yields too few categories, their classification according to speech acts might result in several thousand (Austin 1962: 149). Our communicative sentence types are probably closest to the types of direct illocutionary acts (cf. Searle 1975). It is not always easy to see whether two, segmentally identical but prosodically different utterances are grammatically different (belong to different communicative sentence types) or only attitudinally different (belong to the same communicative sentence type). For instance, we can say that (5)a and (5)b illustrate two distinct communicative sentence types; the former being an exclamation, the latter being a statement (Károly 1964). However, it would be equally justifiable to say that both are statements, but (5)a is “happy” and “surprised”, while (5)b is “neutral”. On the relation between attitudes and communicative sentence types see section 2.3.5. According to Tench (1996: 68), an echo yes–no question “is an identical wording of a part of, or the whole of, a previous utterance, usually for the sake of expressing disbelief or surprise or checking against mishearing.” He also says that strictly speaking such questions only consist of given information but are treated as if they contain the same information structure of given and new as the original utterance which they repeat. Consider, for example, the difficulty of distinguishing exclamations from attitudinally special statements, in connection with examples (5)a and (5)b above.
Notes to Chapter 3: A Taxonomic Analysis of Hungarian Intonation 1. At first sight this kind of taxonomic analysis (based on the modified structuralist maxim “once a prosodeme, always a prosodeme”) may seem somewhat anachronistic, but it is a necessary first step, which was not made for Hungarian when structuralism was at its height. 2. The same analogy is explicit in Gussenhoven’s model of intonation, according to which “an intonation language . . . employs two phonetic resources to encode its morphemes” (Gussenhoven 1985: 117). One subsystem is encoded in spectral composition variations against time (this is the segmental kind of encoding which has always been recognized, giving us the “normal” morphemes of Hungarian), while the other subsystem is encoded in Fo variations against time, and thus contains the “intonational” morphemes of Hungarian. 3. The names “front-falling” and “end-falling” are translations of Deme’s (1962) terms elül eso and végén eso, respectively. 4. The bottom pitch (represented by the horizontal line at the bottom of each intonational diagram) coincides with the pitch of the baseline (as defined by Pierrehumbert 1980: 48), when the latter reaches the end of an intonation phrase; see also Cruttenden (1997: 18).
Notes 205 5. However, when the fall–rise is not utterance-final and is followed by another intonation phrase with a full fall in it, the fall–rise may be an emphatic replacement of the half fall (see section 5.4.7), and is not necessarily ‘conflicting’. 6. The semantic glosses representing the “abstract meanings” of the intonation contours (Cruttenden 1997: 106) are general enough to always be valid, but may receive different pragmatic overtones (“local meanings”, Cruttenden 1997: 91ff) in different contexts. For instance, the general meaning of the fall–rise (“self-contained + conflicting”) may manifest itself in different local meanings, such as “reassuring” (when we want to dispel our partner’s unjustified fear), for example: 7Nem fog fájni! ‘It won’t hurt.’; “warning”, for example: 7Nem alszom! ‘I’m not sleeping.’ (implying: ‘You must believe I am asleep, that’s why you are walking on tiptoe, but I am not asleep.’); “ironic refusal” (when we sarcastically repeat our partner’s words as if we agreed with him/her, though in reality we do not agree with him/her at all), for example: 7Nem forró! ‘It’s not hot.’ (implying: ‘It’s very hot.’); “hopeful yes–no question”, De 7olvastad? ‘But you’ve read it?’ (implying ‘Although this is not obvious, I hope you have read it’); “impatient question-word question”, for example: 7Mikor jönnek? ‘When are they coming?’ (implying ‘You seem to think I am not interested but I am!’). All these are the special local manifestations of the general meaning “self-contained + conflicting”. These facts are in line with Ladd’s opinion: “intonation affects the interpretation of utterances through the interaction of very general meanings and broad principles of pragmatics” (Ladd 1990: 808). My association of several “different” meanings with certain contours in Varga (1981), which Fónagy (1998: 331) partly criticizes, should be seen in this light. 7. The formal and functional differences between the various kinds of end-falling contours were first noticed by Molnár (1954). He related these differences to the magnitude of the angle between the two “wings” of the contour. In our framework, however, these differences are categorical rather than scalar because we distinguish three distinct forms in the left wing: rising, level and descending. 8. It is interesting that the end-falling characters cannot be used in imperatives that require long-lasting or significant activities, for example * ˆOlvasd el ezt a regényt! ‘Read this novel.’ This confirms Fónagy’s hypothesis (1966) that this use of the end-falling contours is a metaphorical extension of their use on yes–no question requests. Yes–no question requests are not used to require long-lasting or significant activities, either; for example ˆElolvasnád ezt a regényt? ‘Would you read this novel?’ cannot be used as a request, only as a real question. 9. This is so even if autosegmentally they may be given the same underlying representation as the yes–no question intonations of some other Eastern European languages (see Grice, Ladd and Arvaniti 2000). For the autosegmental analysis of the Hungarian rise–fall, see Chapter 4. 10. Hungarian learners of English often commit the double mistake of transferring to their English speech the characteristic early stress and rising–falling melody of Hungarian yes–no questions:
206
Notes
la
*
Has Aunt
Ange
come? 11. For special modifications of the basic, two-terraced form of the stylized fall, see Varga (1995). 12. In (22)b there are two major-stressed syllables: kész and va-, and both initiate a stylized fall; for more on this, see section 3.6.1 below; and also section 5.4.4.3. 13. The preparatory contour (to be discussed in section 3.5) is no part of the intonational lexicon because it has no meaning. 14. As has been pointed out in section 3.2.3, if a sustained contour starts with a syllable that has extra intensity, then that syllable is regarded as majorstressed, and the contour as a sustained character contour, rather than as a preparatory contour. 15. However, in a system where no intonation phrases are recognized, a special symbol to mark the beginning of the preparatory contour (for example, [;]) is necessary; see section 3.6.2. 16. The bracketed parts are optional. The structure shown in (32) clearly reminds us of the (Prehead) + (Head) + Nuclear Part division of English tonegroups in traditional British descriptions. The reason why I use different terms for the constituents of the Hungarian IP is that the English terminology suggests that the nuclear part is more prominent than the head. In a Hungarian IP, however, the scale and the terminal part both contain major-stressed syllables, and so the terminal part is not more prominent than the scale. 17. (34) belongs to “quotation-carrying utterances”, described in section 5.3.2, where it will be shown that the two IPs in (34) are in fact two utterances forming a compound utterance. 18. See note 16. 19. As a matter of fact, even English intonation can be, and has been, analysed without establishing IPs, as for example, in Bolinger (1986, 1989). Nevertheless, the general view is that the nucleus should be assigned theoretical status among the English pitch accents and, consequently, that IPs are in fact necessary in English. This is, for example, Ladd’s (1986, 1996) opinion. 20. The phrase Jól ülsz? ‘Are you sitting comfortably?’, with the rising–falling contour of a yes–no question, is often inserted in informal style in a sentence before a surprising constituent. It probably goes back to “Unless you are sitting firmly in your seat, you will fall down with surprise when you hear this.”
Notes to Chapter 4: An Autosegmental Analysis of Hungarian Intonation 1. By “syllable” I mean the traditional unit consisting of an optional onset and an obligatory rhyme, where the rhyme consists of an obligatory nucleus and an optional coda. For an analysis of Hungarian syllables see Törkenczy and Siptár (1999) or Siptár and Törkenczy (2000). For scholars who, like Harris (1994), reject the traditional notion of the syllable but retain syllabic constituents, it is the syllabic nuclei with which the melodic elements of intonation
Notes 207
2. 3. 4. 5.
6.
7.
8. 9. 10.
11. 12.
13. 14. 15. 16. 17.
can be associated. However, it is difficult to see how those newer frameworks which replace syllable strings with CV strings (Lowenstamm 1996) or VC strings (Szigetvári 1999), can be reconciled with the facts of intonation. These frameworks get rid of syllables and syllabic constituents, and the units they offer lack the intonational relevance which the traditional syllables (or their nuclei) have. See our discussion of the intonation of echo and repetitive question-word questions, in section 3.2.4. See Note 4 in Chapter 3. In section 4.5.1 the tone H8 will eventually turn out to be superfluous and so the five levels here distinguished will be reduced to four; see (16). The obligatoriness of the “Obligatory” Contour Principle has also been questioned by Goldsmith (1976: 132–4) and others (van der Hulst and Smith 1985: 16; see also Harris 1994: 172). We shall regard it as obligatory only when there is no reason for preserving the adjacent identical melodic elements. As we shall see, there is good reason for keeping the sequence HH in the case of the monotone-fall and so for ignoring the OCP in this case. The meanings of the basic contours represent the medium degree of some kind of excitement. Applying the Closing Rule to these contours decreases the excitement. The various meanings which the Closing Rule adds to the meanings of the basic contours (that is “finished” for the full fall, “routine” for the descent and “exclaiming” for the descent-fall) can all be seen as the reduction of excitement, and can be generalized as [ – open]; see section 3.2.5. The meaning components added by the Opening Rule (“conflicting” for the fall–rise, “tense” for the rise, and “questioning” for the rise–fall) all increase the excitement inherently present in the meanings of the basic contours and can be generalized as [ + open]; see section 3.2.5. The initial element of the appended contour may be dropped: (L)L9$, especially when it follows a full fall. As has been shown above, H*.H9$ is derived from H*.H$ in the case of the descent, and from H*.L$ in the case of the 2nd-type descent. This list shows that character contours inevitably have at least one H melodic element in their representations, while the appended contour has no H element. The association pattern shown in (28)c may change because the associations in the characters L*.H$ and L*.H.L$ can be rearranged; see section 4.8. The association pattern shown in (29)c will not remain in this form because the association lines in the characters H*.H9$, H*.H9.L$, and H*.H9 + $ will be rearranged; see section 4.8. The association pattern of (30)c may change because the association lines can be rearranged; see section 4.8. The association pattern of (31)c will not remain in this form because the association lines will be rearranged; see section 4.8. The association pattern of (32)c will not remain like this because the association lines will be rearranged; see section 4.8. After a full fall, however, the appended contour may be realized as a monotone at the bottom of the speaker’s voice; see note 8 in this chapter. The third tone of the symmetrical three-tone contours (H*.L.H$ and L*.H.L$) need not be realized at the same height as the first tone. What is
208
18.
19.
20.
21.
Notes important is that it should be higher than the second tone in H*.L.H$, and lower than the second tone in L*.H.L$. On the “gradual”, “end-rising”, and “front-rising” phonetic variants of the rise or the rising part of the rise–fall, see also the relevant diagrams in (5) and (9) in section 3.2 above. According to Grice, Ladd and Arvaniti (2000), the “front-rising” variant of the rising–falling contour is the typical yes–no question intonation of Transylvanian Hungarian. This corresponds to Gussenhoven’s (1985) modification “delay”. Other scholars account for such phenomena by positing a leading tone before the starred one; see, for example, Pierrehumbert (1980), Beckman and Pierrehumbert (1986), Grice (1995). R. Ladd (personal communication) thinks that the use of HL% to indicate sustained non-low final pitch is a specific claim of Pierrehumbert’s analysis of English intonational phonology but need not be carried over to a different language.
Notes to Chapter 5: The Melodic Segmentation of Hungarian Utterances 1. This is actually the second stage in the syntax–phonology mapping, the one following the stage of stress fixing, which will be described in Chapter 6, section 6.6. We assume that the stresses that are relevant in melodic segmentation have already been fixed in the previous stage and are available here. The reason why we are discussing melodic segmentation first is strictly methodological: stress fixing can be described more economically in Chapter 6, that is among the stress-related phenomena to which Part II of the book is devoted. 2. Example (2) is a declarative sentence that carries new information all the way through, that is why each lexical (i.e. non-function) word in it, apart from the verb, receives major stress. The verb is minor-stressed or unstressed when the F position before it is filled, see Chapter 6, sections 6.6.1 and 6.6.3. 3. Vogel (1987) distinguishes seven types of prosodic constituent in Hungarian: the syllable, the foot, the phonological word, the clitic group, the phonological phrase, the intonational phrase, and the phonological utterance (see also Nespor and Vogel 1986, Vogel and Kenesei 1987, Kenesei and Vogel 1989). However, some of these constituent types (the phonological word, the clitic group, the phonological phrase) are not needed for the purposes of our analysis, and the IP is defined differently in our framework (see section 3.6. above). 4. The highest-ranking sentence is not the same as the root sentence. A root sentence is any sentence that is not commanded by a VP node, see Downing (1970: 10). So a co-ordinated sentence like Péter fogorvos, a felesége ügyvéd ‘Peter is a dentist, his wife is a lawyer’ is one highest-ranking sentence, but two root sentences. 5. The term is taken from Hetzron (1980), who defines tonosyntax as that part of syntax which is expressed by intonational means.
Notes 209 6. Observing English examples analogous to Hungarian (8)a, (8)b and (8)c, Ladd (1986) found that, since the continuation after the inserted part shows downdrift, the inserted IP interrupts the matrix IP, so that there are only two IPs, one cutting the other into two discontinuous parts. For Ladd this means that the Strict Layer Hypothesis has to be abandoned. In my view, however, utterances like these can be analysed as having three IPs, and the downdrift observable in the last IP as an instance of inter-IP downdrift, and so the Strict Layer Hypothesis need not be abandoned (see section 3.6.2 above). 7. In (11)b we have one rising–falling character contour, which is interrupted by pauses before and after Angéla, but melodically the contour is undisturbed and constitutes one IP. 8. For an explanation of why the inorganic block here constitutes an appended contour and thus a separate IP rather than a continuation of the full fall that started before it, see section 3.3. 9. This is because a half falling character contour ending with a pause is a terminal contour, see section 3.6. 10. When the head of a constituent is postmodified by a restrictive relative clause, the head must contain, or be premodified by, a demonstrative expression referring to the restrictive relative clause. 11. Heads postmodified by non-restrictive relative clauses, either adjacently or non-adjacently, do not contain a demonstrative expression referring to the subclause. 12. ST6 and ST7 cannot be conflated because they behave differently. When the final block has a fall, the initial block of ST6 can have a sustained contour but the initial block of ST7 cannot; consider examples (49)a and (50)a vs. (51)a and (52)a below. 13. In (32) the major-stressed constituent in F position azért ‘that’s why’ forms a single CB with the unstressed verb feküdt ‘lay-3sg’ and the unstressed N-constituent le ‘down’; see section 5.4.1. 14. In (35) the major-stressed constituent in F position hova ‘where’ forms a single CB unit with the unstressed verb ment ‘went-3sg’. 15. In (37) the major-stressed constituent in F position Pétert ‘Peter-ACC’ forms a single CB with the unstressed verb küldik ‘send-3pl’. 16. The relevant blocks in (42) can also be treated as equivalent blocks, see section 5.4.5 below. 17. See also example (42) above. 18. The example displays other features of emphatic narratives as well. Such a feature is the stressing of words that would not normally be stressed, see Kálmán and Nádasdy (1994: 462–3). For instance, the verb (figyelmeztet), which is not normally major-stressed in post-focal position, is majorstressed here. The high monotone on the pronoun minket is an example of Contour Insertion, to be discussed in Chapter 8.
Notes to Chapter 6: Stress in Hungarian Words, Phrases and Sentences 1. This seems to have changed somewhat in the last few decades. According to Gósy (1997), the quality of vowels is often blurred and approximates to
210
2.
3.
4.
5.
6.
7.
Notes a schwa in spontaneous Hungarian speech. However, since qualitative reduction may occur in both stressed and unstressed positions (even though more often in the latter), it cannot be used for distinguishing stressed syllables from zero-stressed ones. A similar problem occurs in English when it has to be decided whether a post-tonic syllable whose vowel is unreduced is secondary stressed or unstressed, as in the final syllable of anecdote or dynamite. Burzio (1994: 48) complains that, in English, “perceptual evidence does not appear to be a reliable indicator of secondary stress with heavy syllables whose vowel is unreduced. We must therefore rely on other kinds of evidence to determine presence of stress, some necessarily theory-internal, but nonetheless valid.” In English, one kind of such evidence is provided, for example, by the stylized fall, whose second, lower, terrace is initiated by a post-primary secondary stress (see Chapter 3, section 3.2.7). In Hungarian, the possibility of Contour Insertion is a sign of minor stress (see Chapter 8, section 8.3). The only exceptions are a few interjections in which the primary stress falls on the last syllable, as in A[há! ‘Now I see!’, Ho[hó! ‘Oho! Not so fast!’, Apro[pó! ‘By the way! That reminds me!’. Vogel’s (1987) list of prosodic constituent types includes both the foot and the phonological word, and it may at first sight seem that the latter is a more suitable domain of word stress. However, she defines the phonological word as the domain of vowel harmony and n-palatalization, which makes it irrelevant in our framework and so it need not concern us. This does not mean that it is unnecessary; it is simply a constituent which we do not have to take into consideration. The objection that phrasal compounds are phrases rather than compounds can be refuted on grounds that the constituent words within these items have grown together through their constant co-existence (have been lexicalized into compounds). The existence of phrasal compounds is also recognized in English, see, for example, Fudge (1984: 144ff), Cinque (1993: 280ff). For instance, [Christmas tree is an ordinary compound, [Christmas [pudding is a phrasal one. The changes that the stress pattern of a phrasal compound can undergo when it is embedded in a phrase will be described as rhythmical variation in Chapter 7. In ordinary Hungarian compounds, even a branching second constituent is weaker than the first. Both kerékcsapágy in (11)a and csapágy in (11)b are branching contituents but nevertheless weaker than the first constituent in their respective words. By contrast, in an English compound, a branching second constituent is said to be stronger than the first, see for example, law degree language requirement; see Liberman and Prince (1977: 257ff), Cinque (1993: 273ff). In a framework using grids, the equivalent of Lexical Foot Erasing would be Line Conflation, as Halle and Vergnaud (1987: 52) define it. LC means replacement of the material on Line 1 by the corresponding material on Line 2. (For Halle and Vergnand the bottom line is Line 0.) For instance:
Line Line Line
2 1 0
x x x x x
3
x x x x
Notes 211 8. The compound word is a lexical unit and not a separate prosodic unit. Since its prosody develops within the lexical stratum of phonology, I do not subscribe to Vogel’s (1987) opinion that compounds belong to clitic groups, because the latter are formed in the postlexical stratum. 9. However, the speaker may deliberately suspend the otherwise obligatory Clash Deletion rule, and retain the minor stress if there is a need for disambiguation. 10. Hunyadi (1999) goes as far as saying that in some cases the minor stresses cannot be deleted at all, which, if correct, means that our UFE may be blocked. He would not allow the deletion of the minor stresses of an operator + scope complex which is embedded in the reduced scope of an operator, as, for example, in the following: [Nem eheted meg csak a ]torta ]tetejét. ‘You are NOT allowed to eat just the top of the cake.’ (Hunyadi 1999: 161). 11. The Hungarian definite article has the form a when the word following it begins with a consonant, and az when the word following it begins with a vowel. 12. Those monosyllabic function words that are contrasted or focused will be major-stressed, of course, and do not cliticize. 13. According to Vogel (1987) the prosodic constituent containing a non-clitic lexical item with or without clitics before and/or after it is the clitic group, which is the domain of a single primary stress. In a left-end based framework of feet, however, the clitic group can be dispensed with. 14. This is different from the case in which the function word is major-stressed for focus or contrast. 15. In rejecting the idea that non-clitic function words (including pronouns) are unstressed, I agree with Selkirk (1984: 342): “making the opposite assumption – that function words are basically unstressed (and thus that their stressed forms are derived by rule in certain specified circumstances) – requires considerable ad hoc complication in the grammar.” 16. By contrast, in English phrases and sentences conveying new information, the last stress tends to be noticeably stronger than the other stresses before it, due to the English Nuclear Stress Rule. 17. According to É. Kiss (1987–88), the topic of a sentence is typically minorstressed before the major-stressed intitial constituent of the comment. In order to achieve minor stress on the topic, É. Kiss (1987–88: 12) proposes a filter which eliminates any pitch accent on it. In reality, however, in far too many cases the topic does carry a pitch accent, and so it is major-stressed. This major stress may be less strong than the one at the beginning of the comment, but this is irrelevant in our framework. Besides, the two may also be equally strong (Fónagy 1998: 340). In more recent work É. Kiss (1992: 95) herself admits this possibility. 18. According to some views (for example, Vogel 1987: 246), the verb is not only deaccented (i.e. minor-stressed) but also destressed (i.e. zero-stressed) after a filled F position. I do not agree with this because the first syllable of the verb may become the target of Contour Insertion (see Chapter 8): | [Kávét 4iszik a | [földszinti 1büfében |, and so the verb must be minor-stressed. The verb is completely destressed when the F constituent before it is monosyllabic; this destressing is the result of Clash Deletion (see section 6.6.3).
212
Notes
19. The verb becomes zero-stressed here, due to Clash Deletion, to be discussed in section 6.6.3. 20. When the verb has no reduced complement, as in (i), or when the verb is accompanied by a negative or imperative operator, as in (ii) and (iii), respectively, the F can remain unfilled (cf. É. Kiss 1981):
(i)
(ii)
(iii)
V[[Hívtam]
a [barátodat. called-1sg the friend-your-ACC ‘I called your friend.’ V[[Nem hívtam] a barátodat. not ‘I did not call your friend.’ a barátodat! V[[Hívd] call-2sg-IMP ‘Call your friend.’
21. The reduced complement must move from the F position not only when the F position is filled by an emphatic complement, but also when it is filled by a complement modified by a “focusing” operator (i.e. a negative, interrogative, optative, or exclamative operator, the particle csak ‘only’, etc.); see É. Kiss (1981). In all these cases the F position lends major stress to the constituent if it does not bring major stress with it:
(i)
Mi F[ot] hívtuk meg. ‘It is him that we invited.’ we him called PERF
22. É. Kiss (1992: 167–71) explains sentences like (35)a as stylistic versions of sentences with the quantifier being in Q position, for example [Sokan [nem beszéltek. For a detailed account of how operator and scope are prosodically expressed, see Hunyadi (1999). 23. Selkirk (1986), based on her suggestion for end-based mapping rules for prosodic constituents, regards this domain as a prosodic constituent: viz. the phonological phrase. This “phonological phrase” is different from Vogel’s (1987). The latter groups together a clitic group containing a lexical head X with all clitic groups on its non-recursive (i.e. right) side. So complements or modifiers on the left side of a head (as, for example, the word földszinti in a földszinti büfében ‘in the groundfloor snackbar’) can only be accommodated into a common phonological phrase with the head büfében by means of restructuring. This problem does not arise under Selkirk’s definition, which is therefore preferable. However, in our framework, in which the syntactic S-structure of the sentence is present in the course of the syntax–phonology mapping, the notion of the phonological phrase (in either Selkirk’s or Vogel’s sense) can be dispensed with.
Notes to Chapter 7: Rhythmical Variation in Phrasal Compounds 1. The term Trochaic Reversal is used as the opposite of Iambic Reversal, which is the name of the English counterpart of the process. It was first used by
Notes 213
2. 3.
4.
5.
Gussenhoven (1983b) for one of the rhythm rules of Dutch. However, Gussenhoven’s Trochaic Reversal is the durational restructuring of postaccentual syllables and does not involve changes of intonational pitch accents. In Varga (1998) I called this Final Weakening. If the numerals in (5)a–b are followed by any of the words ezer ‘thousand’ or millió ‘million’ or milliárd ‘billion’, they remain double-accented and RV, for example [tizen[háromezer ‘13,000’ or [tizen[három millió ‘13,000,000’ or [tizen[három milliárd ‘13,000,000,000’. This is because the words ezer ‘thousand’, millió ‘million’, milliárd ‘billion’, (as well as the word száz ‘hundred’), when preceded by numerals in numeric compounds, do not receive a separate major stress, for example: [ötezer ‘5,000’, [tízmillió ‘10,000,000’; [hatszáz ‘600’. As will be shown in section 7.2.2, Trochaic Reversal is blocked if its application leads to more than four interaccentual syllables. This considerably limits the possibilities of producing derivatively double-accented numeric compounds. For instance [kétszáz]tizen[hat ‘216’ is a derivatively doubleaccented numeral, produced by applying Trochaic Reversal to tizenhat. However, [kétszáz ]tizen[hat can hardly undergo Trochaic Reversal in [ezer[két száz]tizen[hat ‘1216’ to produce a double-accented *[ezer]kétszáz]tizen[hat because the output would contain five syllables between the two remaining accents. It seems that in addition to the RV numeric and ugyan-compounds there are a few stereotyped collocations which may also occasionally display rhythmical variation. These include semi-lexicalized phrases, for example, [két [tojás ‘two eggs’, [egész [nap ‘all day’, and double surnames or double first names within three-word names, for example, [Gór [Nagy [Mária, [Lázár [Armand [Péter.
(i)
a. b. (ii) a. b. (iii) a. b.
[Végy két [tojást. ‘Take [Két tojás [sárgája. ‘The [Aludt ]egész [nap. ‘He [Egész ]nap [aludt. ‘The [Gór Nagy [Mária [Lázár ]Armand [Péter
two eggs.’ yolk of two eggs.’ was sleeping the whole day.’ whole day he was sleeping.’
However, collocations like these are sporadic, and they undergo rhythmical variation in a less systematic fashion than the double-accented numeric and ugyan-compounds do, so we do not deal with them here. 6. The syntactically proper analysis of year numbers is right-branching: [ezer[kilencszáz[ötven[hat. [E [A [B C] ] ] However, this analysis would only allow the reduction of major stress over the syllable öt-: [ezer[kilencszáz]ötven[hat. The fact that Hungarian native speakers prefer [ezer]kilencszáz[ötven[hat or [ezer]kilencszáz]ötven[hat shows that a re-analysis of year numbers as [[B1C1] [B2C2]] may take place. 7. In Varga (1998) I referred to this as Initial Weakening. 8. Hungarian shows a relative constancy in the duration of syllables and a lack of vowel reduction, so each syllable has to be aligned not only with a demibeat on Level 1 (L1) but also with a basic beat on Level 2 (L2) (Selkirk 1984: 41).
214
9.
10. 11.
12.
Notes These are features of syllable-timing, even though Hungarian is not a pure syllable-timed language. Since in Hungarian every syllable has an x on both L1 and L2, we could simplify our grids by leaving out L1 and starting with L2 as the bottom level. However, this could lead to confusion, and so it will not be done. This condition says that if, by the application of Beat Addition, another syllable reaches the prominence level of the originally strongest stressed syllable of the phrase, then the originally strongest stressed syllable will be automatically promoted by one level. For derivations based on the Hungarian Rhythm Rule, see Varga (1998: 252–4). In English, the corresponding two types of rhythmical variation, Iambic Reversal and Iambic Confirmation, are very general in their application and affect the same kinds of words (viz. words having two major stresses, not counting the lexical exceptions), and both types can be regarded as essentially belonging in stratum P1 of postlexical phonology. If Trochaic Confirmation does not take place, the form [tizen[három ]negyven[öt becomes the output version.
Notes to Chapter 8: Rhythmical Secondary Stresses 1. In the words of Maranungku (a language spoken in Australia) “[p]rimary stress falls on the initial syllable, and secondary stress falls on every other syllable thereafter” (Kenstowicz, 1994: 555). 2. Szinnyei’s reference to the lightness of the third syllable is quite unique in the Hungarian linguistic literature, where syllabic weight is not normally regarded as having any relevance for phonological rules (though see Gráf 2001). 3. It can be pointed out that Contour Insertion is similar to Hayes’ (1995) diagnostics for secondary stress in English: a pitch accent is associated with a secondary stressed syllable. However, there is a difference. In Hayes (1995: 17), a rising contour (or rather its L* accent) “seeks out the vowel with secondary stress in preference to stressless vowels” but the secondary stress taking that contour remains a secondary stress. In Hungarian this cannot happen, because where pitch accent (i.e. intonational contour initiation) and extra intensity meet, we get a major stress. If a high monotone (or rise) is associated with a syllable which would be minor-stressed without it, the syllable becomes major-stressed (see section 3.2.1). In Contour Insertion the syllable with the newly introduced high monotone is not weaker than the syllable with the full fall after it. 4. There was only one exception to this: (10)d, which I included to check whether or not a stress clash mattered in the evaluation. This variant was received extremely unfavourably and obtained the lowest score in its block. 5. For the statistical analysis I am indebted to Anita Csölle, of Eötvös Loránd University, Budapest. 6. The fact that the rejected variants also scored points shows that (a) listeners were unable to make absolute judgements about such subtle differences; and that (b) even the rejected variants had been rendered in a natural enough
Notes 215 way, so as not to bias the participants against them, which demonstrates the reliability of the experiment. In any case, these score numbers show why the statistical significance calculations are especially important here. 7. In English, a pre-primary secondary stress may trigger Iambic Reversal (Liberman and Prince 1977: 319), and a post-primary secondary stress (actually, the first one of the post-primary secondary stresses) initiates the lower terrace of the Stylized Fall (Gussenhoven 1985: 124). 8. Secondary stresses in Spanish have also been shown to be postlexical (Roca 1986). Roca (1986: 341) claims that the domain of Spanish secondary stress is the “phonic group”, which he does not define but says is parallel to Dell’s “tronçon”. Since this is “any stretch of utterance comprising no silence or intonational break” (Roca 1986: 365, note 2), it seems to be very close to our intonational phrase.
References
Antipova, A. M. (1979) Sistema anglijskoj rechevoj intonacii [The system of English speech intonation] (Moskva: Vysshaja shkola). Austin, J. L. (1962) How to Do Things with Words (Oxford: Clarendon Press). Balassa, J. (1887–90) Hangsúly a magyar nyelvben. [Stress in the Hungarian language.] Nyelvtudományi Közlemények, 21: 401–34. Beckman, M. E. and Pierrehumbert, J. B. (1986) Intonational structure in English and Japanese. Phonology Yearbook, 3: 255–309. Bing, J. (1980) Aspects of English Prosody (Bloomington: Indiana University Linguistics Club). Bolinger, D. L. (1957) Interrogative structures of American English (The direct question). Publication of the American Dialect Society, 28 (University of Alabama Press). Bolinger, D. L. (1957–8) Intonation and grammar. Language Learning, 8: 31–8. Bolinger, D. L. (1964) Around the edge of language: intonation. Harvard Educational Review, 34: 282–96. Bolinger, D. L. (1972; 1st publ. 1970) Relative height. In: D. L. Bolinger (ed.), Intonation (Harmondsworth: Penguin) pp. 137–53. Bolinger, D. L. (1978) Intonation across languages. In: J. H. Greenberg (ed.), Universals of Human Language, vol. 2: Phonology (Stanford, CA: Stanford University Press) pp. 471–524. Bolinger, D. L. (1986) Intonation and its Parts (London: Edward Arnold). Bolinger, D. L. (1989) Intonation and its Uses (London: Edward Arnold). Bolla, K. (1980) A fonetikai szerkezetek interlingvális egybevetésérol. (Problémavázlat.) [On the interlingual comparison of phonetic structures (A Sketch of the problem).] Magyar Fonetikai Füzetek, 5: 40–69. Brassai, S. (1888) Szórend és accentus. [Word order and accent.] Értekezések a Nyelv-és Széptudományi Osztályok körébol, 14/9. Brassai, S. (1889) Szerény észrevételek a “Hangsúly” címu értekezés irányában. [Modest observations concerning the treatise “Stress”.] Magyar Nyelvor, 18: 341–51, 396–406. Brazil, D. (1985) The Communicative Value of Intonation in English, Discourse Analysis Monograph No. 8 (Birmingham: Bleak House and English Language Research). Brazil, D. (1995) A Grammar of Speech (Oxford: Oxford University Press). Brazil, D., Coulthard, M. and Johns, C. (1980) Discourse Intonation and Language Teaching (London: Longman). Bryzgunova, E. A. (1971) O smyslorazlichitelnyh vozmozhnostjah russkoj intonacii. [On the sense-distinguishing possibilities of Russian intonation.] Voprosy Jazykoznanija, 52: 42–52. Burzio, L. (1994) Principles of English Stress (Cambridge: Cambridge University Press). Cheremisina, N.V. (1982) Russkaja intonacia: poezia, proza, razgovornaja rech [Russian intonation: poetry, prose, conversational speech] (Moskva: Russkij jazyk). 216
References
217
Cinque, G. (1993) A null theory of phrase and compound stress. Linguistic Inquiry, 24: 239–97. Coates, J. (1975) Attitudinal factors in Hungarian and English intonation. Idegen Nyelvek Tanítása, 18: 52–6. Cruttenden, A. (1997) Intonation, 2nd edn (Cambridge: Cambridge University Press). Crystal, D. (1969) Prosodic Systems and Intonation in English (Cambridge: Cambridge University Press). Csury, B. (1925) A szamosháti nyelvjárás hanglejtésformái. [The intonation patterns of the dialect of Szamoshát.] A Magyar Nyelvtudományi Társaság Közleményei, 22. Csury, B. (1935) A mondathanglejtés. [Sentence intonation.] Debreceni Szemle, 9: 105–19. Csury, B. (1939) A magyar kiejtés kérdése. [The issue of Hungarian pronunciation.] Magyar Nyelv, 35: 32–9. Daneš, F. (1960) Sentence intonation from a functional point of view. Word, 16: 34–54. Deme, L. (1962) Hangsúly, szórend, hanglejtés, szünet. [Stress, word order, intonation, pause.] In: J. Tompa (ed.), A mai magyar nyelv rendszere [The system of present-day Hungarian] vol. II (Budapest: Akadémiai Kiadó) pp. 457–522. Deme, L. (1970) A kiejtés törvényeinek tanítása és tanulmányozása. [Teaching and studying the rules of pronunciation.] Magyar Nyelvor, 89: 188–200. Deme, L. (1979) Grammatikai képlet és akusztikai képlet kapcsolatához. [On the connection between grammatical pattern and acoustic pattern.] Magyar Fonetikai Füzetek, 3: 7–13. Downing, B. (1970) Syntactic structure and phonological phrasing in English. Ph.D. dissertation, University of Texas. Elekfi, L. (1964) Az aktuális mondattagolás egyik alapformája a magyarban. [One of the basic forms of actual sentence bipartition in Hungarian.] Nyelvtudományi Közlemények, 66: 331–70. Elekfi, L. (1968) Mondatformák és mondathangsúlyok fonetikai vetülete. [The phonetic reflexes of sentence forms and sentence stresses.] Nyelvtudományi Közlemények, 70: 329–50. É. Kiss, K. (1981) Structural relations in Hungarian: a “free” word order language. Linguistic Inquiry, 12: 185–213. É. Kiss, K. (1987) Configurationality in Hungarian (Dordrecht: Reidel and Budapest: Akadémiai Kiadó). É. Kiss, K. (1987–88) Még egyszer a magyar mondat intonációjáról és hangsúlyozásáról. [Once more on the intonation and stressing of the Hungarian sentence.] Nyelvtudományi Közlemények, 89: 1–52. É. Kiss, K. (1992) Az egyszeru mondat szerkezete. [Structure of the simple sentence.] In: F. Kiefer (ed.), Strukturális magyar nyelvtan, vol. 1: Mondattan [A structural grammar of Hungarian, vol. 1: Syntax] (Budapest: Akadémiai Kiadó) pp. 79–177. Fogarasi, J. (1838) Eυρηκα [Athenaeum], XIII, XIV, XIX. Fónagy, I. (1966) Api, figyelj ide!? [Daddy, listen!] Magyar Nyelvor, 90: 121–37. Fónagy, I. (1989) On the status and functions of intonation. Acta Linguistica Hungarica, 39: 53–92. Fónagy, I. (1998) Intonation in Hungarian. In: D. Hirst and A. Di Cristo (eds), Intonation Systems: A Survey of Twenty Languages (Cambridge: Cambridge University Press) pp. 328–44.
218
References
Fónagy, I. and Magdics, K. (1967) A magyar beszéd dallama [The melody of Hungarian speech] (Budapest: Akadémiai Kiadó). Fudge, E. C. (1984) English Word-Stress (London: George Allen and Unwin). Goldsmith, J. A. (1976) Autosegmental Phonology (Bloomington: Indiana University Linguistics Club; published by Garland Press, 1979). Goldsmith, J. A. (1982) Review of The Structure of Intonational Meaning: Evidence from English by D. Robert Ladd, Jr (Bloomington: Indiana University Press, 1980), in Language, 58: 422–4. Goldsmith, J. A. (1990) Autosegmental and Metrical Phonology (Oxford: Blackwell). Gósy, M. (1997) A magánhangzók minoségváltozásai a spontán beszédben. [Qualitative changes of vowels in spontaneous speech.] Magyar Nyelvor, 121: 45–56. Gráf, Z. B. (2001) Length in metrical phonology. Ph.D. dissertation, Eötvös Loránd University, Budapest. Greenberg, S. R. (1969) An experimental study of certain intonation contrasts in American English. Working Papers in Phonetics 13 (Los Angeles, CA: University of California). Grice, M. (1995) Leading tones and downstep in English. Phonology, 12: 183–223. Grice, M., Ladd, D. R. and Arvaniti, A. (2000) On the place of phrase accents in intonational phonology. Phonology, 17: 143–85. Gumperz, J. J. and Kaltman, H. (1980) Prosody, linguistic diffusion and conversational inference. Proceedings of the 6th Annual Meeting of the Berkeley Linguistics Society (Berkeley: University of California) pp. 44–65. Gussenhoven, C. (1983a) A Semantic Analysis of the Nuclear Tones of English (Bloomington: Indiana University Linguistics Club). Gussenhoven, C. (1983b) Stress shift and the nucleus. Linguistics, 21: 303–39. Gussenhoven, C. (1985) Intonation: a whole autosegmental language. In: H. van der Hulst and N. Smith (eds), Advances in Nonlinear Phonology (Dordrecht and Cinnaminson: Foris) pp. 117–31. Gussenhoven, C. (1991) The English rhythm rule as an accent deletion rule. Phonology, 8: 1–35. Hall, R. A., Jr. (1944) Hungarian grammar. Language Monograph No. 21. Supplement to Language, vol. 20, no. 4. Halle, M. (1998) The stress of English words, 1968–1998. Linguistic Inquiry, 29: 539–68. Halle, M. and Idsardi, W. (1995) General properties of stress and metrical structure. In J. A. Goldsmith (ed.), The Handbook of Phonological Theory (Cambridge, MA and Oxford: Blackwell) pp. 403–43. Halle, M. and Vergnaud, J.-R. (1987) An Essay on Stress (Cambridge, MA and London: MIT Press). Halliday, M. A. K. (1967) Intonation and Grammar in British English (The Hague: Mouton). Hammond, M. (1987) Hungarian cola. Phonology Yearbook, 4: 267–9. Hammond, M. (1999) The Phonology of English: A Prosodic Optimality-Theoretic Approach (Oxford: Oxford University Press). Harris, J. (1994) English Sound Structure (Oxford and Cambridge, MA: Blackwell). Hayes, B. (1984) The phonology of rhythm in English. Linguistic Inquiry, 15: 33–74. Hayes, B. (1990) Precompiled phrasal phonology. In: S. Inkelas and D. Zec (eds), The Phonology–Syntax Connection (Chicago and London: University of Chicago Press) pp. 85–108.
References
219
Hayes, B. (1995) Metrical Stress Theory: Principles and Case Studies (Chicago and London: University of Chicago Press). Hetzron, R. (1980) Ízelíto a magyar tonoszintaxisból. [A Taste of Hungarian Tonosyntax.] Nyelvtudományi Értekezések, 104: 389–98. Hetzron, R. (1992) Prosodic morphemes in Hungarian. In: I. Kenesei and Cs. Pléh (eds), Approaches to Hungarian 4: The Structure of Hungarian (Szeged: JATE) pp. 141–56. van der Hulst, H. and Smith, N. (1985) The framework of nonlinear generative phonology. In: H. van der Hulst and N. Smith (eds), Advances in Nonlinear Phonology (Dordrecht and Cinnaminson: Foris) pp. 3–55. Hunyadi, L. (1999) Hungarian Sentence Prosody and Universal Grammar. PostPh.D. dissertation, Debrecen. Inkelas, S. and Leben, W. R. (1987) The phonology of intonation in Hausa. In: W. Poser (ed.), Phonetics–Phonology Interface (Course material for the 1987 Linguistic Institute, Stanford University, Stanford, CA) pp. 123–38. Joannovics, Gy. (1889) Hangsúly. [Stress.] Magyar Nyelvor, 18: 61–8, 100–13, 153–60, 200–12, 249–62. Jones, D. (1967; 1st edn. 1918) An Outline of English Phonetics, 9th edn (Cambridge: Heffer). Juhász, F. (1968) Stress and Intonation as Construction Markers in Hungarian. Ph.D. dissertation, Columbia University. Kager, R. (1995) The metrical theory of word stress. In: J. A. Goldsmith (ed.), The Handbook of Phonological Theory (Cambridge, MA and Oxford: Blackwell) pp. 367–402. Kager, R. and Visch, E. (1988) Metrical constituency and rhythmic adjustment. Phonology, 5: 21–71. Kaisse, E. M. (1985) Connected Speech (Orlando: Academic Press). Kaisse, E. M. (1990) Toward a typology of postlexical rules. In: S. Inkelas and D. Zec (eds), The Phonology–Syntax Connection (Chicago and London: The University of Chicago Press) pp. 127–43. Kálmán, L. and Kornai, A. (1985) Hogy intonál a magyar? [How do Hungarians use intonation?] Nyelvtudományi Közlemények, 87: 293–310. Kálmán, L. and Nádasdy, Á. (1994) A hangsúly. [Stress.] In: F. Kiefer (ed.), Strukturális magyar nyelvtan, vol. 2: Fonológia [A Structural Grammar of Hungarian, vol. 2: Phonology] (Budapest: Akadémiai Kiadó) pp. 393–467. Kálmán, C. Gy., Kálmán, L., Nádasdy, Á. and Prószéky, G. (1989) A magyar segédigék rendszere. [The system of Hungarian auxiliary verbs.] Általános Nyelvészeti Tanulmányok, 17: 49–103. Károly, S. (1964) A mondatfajták vizsgálata a funkció és forma szempontjából. [An investigation of sentence types from the point of view of function and form.] Nyelvtudományi Közlemények, 66: 67–88. Kassai, I. (1979) Az intonáció szerepe a gyermeknyelvben. [The role of intonation in child language.] Magyar Fonetikai Füzetek, 3: 80–3. Kenesei, I. (1998) Adjuncts and arguments in focus. Acta Linguistica Hungarica, 45: 61–88. Kenesei, I. and Vogel, I. (1989) Prosodic phonology in Hungarian. Acta Linguistica Hungarica, 39: 149–93. Kenstowicz, M. (1994) Phonology in Generative Grammar (Cambridge, MA and Oxford: Blackwell).
220
References
Kerek, A. (1971) Hungarian metrics (some linguistic aspects of iambic verse). Indiana University Publications Uralic and Altaic Series, vol. 117. Kicska, E. (1891) Hangsúly és szórend. [Stress and word order.] Magyar Nyelvor, 20: 292–8, 337–45, 385–94, 433–45, 481–91. Kiefer, F. (1967) On emphasis and word order in Hungarian. Indiana University Publications, Uralic and Altaic Series, vol. 76. Kiefer, F. (1983) Az elofeltevések elmélete [Theory of Presuppositions] (Budapest: Akadémiai Kiadó). Kontra, M. (ed.) (1988) Beszélt nyelvi tanulmányok [Studies in Spoken Language.] Linguistica, Series A: Studia et Dissertationes, 1. Budapest: Institutum Linguisticum Academiae Scientiarum Hungaricae. Kornai, A. and Kálmán, L. (1989) Hungarian sentence intonation. In: H. van der Hulst and N. Smith (eds), Autosegmental Studies on Pitch Accent (Dordrecht: Foris) pp. 183–95. Kozma, E. (1974) A magyar intonáció funkcionális vizsgálatának újabb lehetoségei Brizgunova elmélete alapján. [New possibilities of the functional examination of Hungarian intonation on the basis of Bryzgunova’s theory.] Magyar Nyelvor, 98: 476–88. Kozma, E. and Szende, T. (1981) Elmenni elmehetek . . . ?! Egy önálló szövegfonológiai dallamtípus elemzése. [Go I can . . . ?! Analysis of an independent melodic type of text-phonology.] Magyar Nyelv, 77: 141–8. Ladd, D. R. (1980) The Structure of Intonational Meaning (Bloomington and London: Indiana University Press). Ladd, D. R. (1983) Phonological features of intonational peaks. Language, 59: 721–59. Ladd, D. R. (1986) Intonational phrasing: the case for recursive prosodic structure. Phonology Yearbook, 3: 311–40. Ladd, D. R. (1990) Intonation: emotion vs. grammar. Review of Intonation and Its Uses by Dwight Bolinger. Language, 66: 806–16. Ladd, D. R. (1996) Intonational Phonology (Cambridge: Cambridge University Press). Laziczius, Gy. (1944) Fonétika [Phonetics] (Budapest: Királyi Magyar Egyetemi Nyomda). Leben, W. (1973) Suprasegmental Phonology (Bloomington: Indiana University Linguistics Club). Leben, W. (1976) The tones of English intonation. Linguistic Analysis, 2: 69–107. Leech, G. (1974) Semantics (Harmondsworth: Penguin). Lehiste, I. (1970) Suprasegmentals (Cambridge, MA: The MIT Press). Liberman, M. (1978) The Intonational System of English. Ph.D. dissertation, MIT. Distributed by Indiana University Linguistics Club, Bloomington. Liberman, M. and Prince, A. (1977) On stress and linguistic rhythm. Linguistic Inquiry, 8: 249–336. Liberman, M. and Sag, I. (1974) Prosodic form and discourse function. Papers from the 10th Regional Meeting of the Chicago Linguistic Society, 416–27. Lowenstamm, J. (1996) CV as the only syllable type. In: J. Durand and B. Laks (eds), Current Trends in Phonology: Models and Methods (European Studies Research Institute: University of Salford Publications) pp. 419–42. Molnár, I. (1954) A magyar hanglejtés rendszere – A magyar énekbeszéd recsitativóban és ariozóban [The System of Hungarian Intonation – Hungarian SongSpeech in Recitative and Arioso] (Budapest: Zenemukiadó).
References
221
Nádasdy, Á. (1985) Segmental phonology and morphophonology. In: I. Kenesei (ed.), Approaches to Hungarian, vol. 1: Data and Descriptions (Szeged: JATE) pp. 225–46. Nádasdy, Á. (1995) Review of Miklós Kontra (ed.), Beszélt nyelvi tanulmányok [Studies in Spoken Language]. Linguistica, Series A: Studia et Dissertationes, 1. Budapest: Institutum Linguisticum Academiae Scientiarum Hungaricae (1988) International Journal of the Sociology of Language, 111: 103–06. Németh, T. Eniko (1996) A szóbeli diskurzusok megnyilatkozáspéldányokra tagolása. [Division of spoken discourse into utterance tokens.] Nyelvtudományi Értekezések, 142. Nespor, M. and Vogel, I. (1986) Prosodic Phonology (Dordrecht: Foris). O’Connor, J. D. (1973) Phonetics (Harmondsworth: Penguin). O’Connor, J. D. and Arnold, G. F. (1961) Intonation of Colloquial English (London: Longman). Papp, I. (1966) Leíró magyar hangtan [Descriptive Hungarian phonetics] (Budapest: Tankönyvkiadó). Peshkovskij, A. M. (1959; 1st publ. 1928) Intonacia i grammatika. [Intonation and grammar.] In: Izbrannie trudi [Selected works] (Moskva: Gosudarstvennoje Uchebno-pedagogicheskoje Izdatelstvo Ministerstva Prosveschenia RSFSR) pp. 171–91. Péter, M. (1961) O nekotoryh voprosah intonacii. (Na materiale russkovo jazyka.) [On certain questions of intonation. (Based on Russian linguistic material).] Annales Universitatis Scientiarum Budapestinensis de Rolando Eötvös nominatae, Sectio philologica, 3: 131–42. Péter, M. (1984) Érzelemkifejezés, stílusérték és expresszivitás a nyelvben. [Emotional expression, stylistic value and expressivity in language.] Általános Nyelvészeti Tanulmányok, 15: 219–35. Pierrehumbert, J. B. (1980) The Phonetics and Phonology of English Intonation. Ph.D. dissertation, MIT. Pike, K. (1945) The Intonation of American English (Ann Arbor: University of Michigan Press). Pope, E. (1971) Answers to yes–no questions. Linguistic Inquiry, 2: 69–82. Roach, P. (1983) English Phonetics and Phonology: A Practical Course (Cambridge: Cambridge University Press). Roca, I. (1986) Secondary stress and metrical rhythm. Phonology Yearbook, 3: 341–70. Roca, I. and Johnson, W. (1999) A Course in Phonology (Oxford and Malden, MA: Blackwell). Searle, J. (1975) Indirect speech acts. In: P. Cole and J. Morgan (eds), Syntax and Semantics, vol. 3: Speech Acts (New York: Academic Press) pp. 59–82. Selkirk, E. O. (1984) Phonology and Syntax: The Relation between Sound and Structure (Cambridge, MA: MIT Press). Selkirk, E. O. (1986) On derived domains in sentence phonology. Phonology Yearbook, 3: 371–405. Simonyi, Zs. (1880) Az “is” szórendi szerepe. [The role of “is” < ‘also’ > in the order of words.] Magyar Nyelvor, 9: 289–99, 337–45. Siptár, P. and Törkenczy, M. (2000) The Phonology of Hungarian (Oxford: Oxford University Press). Szabolcsi, A. (1981) Compositionality in focus. Folia Linguistica, 15: 141–61.
222
References
Szende, T. (1976) A beszédfolyamat alaptényezoi [Basic factors in the speech flow] (Budapest: Akadémiai Kiadó). Szigetvári, P. (1999) VC Phonology: a theory of consonant lenition and phonotactics. Ph.D. dissertation, Budapest: Eötvös Loránd University. Szinnyei, J. (1912) Ungarische Sprachlehre (Berlin: Göschen). Tench, P. (1996) The Intonation Systems of English (London: Cassell). Tolnai, V. (1915) Adatok a magyar hanglejtéshez. [Data for Hungarian intonation.] Magyar Nyelv, 11: 51–9, 108–16, 152–6. Törkenczy, M. and Siptár, P. (1999) Hungarian syllable structure: arguments for/ against complex constituents. In: H. van der Hulst and N. A. Ritter (eds), The Syllable, Views and Facts (Berlin: Mouton & de Gruyter) pp. 249–84. Torsujeva, I.G. (1979) Intonacia i smysl vyskazyvania [Intonation and the meaning of the utterance] (Moskva: Nauka). Uldall, E. T. (1960) Attitudinal meanings conveyed by intonation. Language and Speech, 3: 223–34. Uldall, E. T. (1964) Dimensions of meaning in intonation. In: D. Abercrombie et al. (eds), In Honour of Daniel Jones: Papers Contributed on the Occasion of his Eightieth Birthday (London: Longman Green) pp. 271–9. Varga, L. (1981) A magyar intonáció – funkcionális szempontból. [Hungarian intonation – from a functional point of view.] Nyelvtudományi Közlemények, 83: 313–39. Varga, L. (1983) Hungarian sentence prosody: an outline. Folia Linguistica, 17: 117–51. Varga, L. (1985) Intonation in the Hungarian sentence. In: I. Kenesei (ed.), Approaches to Hungarian, vol. 1: Data and Descriptions (Szeged: JATE) pp. 205–24. Varga, L. (1986) Vélemények a magyar mondat hangsúlyozásáról – avagy Brassai és a többiek. [Opinions on the stressing of the Hungarian sentence – or Brassai and the others.] Nyelvtudományi Közlemények, 88: 181–8. Varga, L. (1987–88) Hozzászólás egy hangsúlytanulmányhoz. [Remarks on a study of stress.] Nyelvtudományi Közlemények, 89: 53–66. Varga, L. (1989) A beszéddallamok fonológiai, szemantikai és szintaktikai vonatkozásai (Magyar anyag alapján). [The phonological, semantic and syntactic aspects of speech melodies (On the basis of Hungarian).] Budapest: CSc dissertation [Ph.D. administered by the Hungarian Academy of Sciences]. Varga, L. (1993) A magyar beszéddallamok fonológiai, szemantikai és szintaktikai vonatkozásai. [The phonological, semantic and syntactic aspects of Hungarian speech melodies.] Nyelvtudományi Értekezések, 135. Varga, L. (1994) A hanglejtés. [Intonation]. In: F. Kiefer (ed.), Strukturális magyar nyelvtan, vol. 2: Fonológia [A Structural Grammar of Hungarian, vol. 2: Phonology] (Budapest: Akadémiai Kiadó) pp. 468–549. Varga, L. (1995) Stylization of the falling tone in Hungarian intonation. In: J. Windsor Lewis (ed.), Studies in General and English Phonetics, Essays in Honour of Professor J. D. O’Connor (London: Routledge) pp. 278–87. Varga, L. (1998) Rhythmical variation in Hungarian. Phonology, 15: 227–66. Varga, L. (2000) A magyar mellékhangsúly fonológiai státusáról [On the phonological status of secondary stress in Hungarian.] Magyar Nyelvor, 124: 91–108. Vértes, O. A. (1980) A magyar leíró hangtan története az újgrammatikusokig [History of Hungarian descriptive phonetics till the Young Grammarians] (Budapest: Akadémiai Kiadó).
References
223
Vogel, I. (1987) Prosodic constituents in Hungarian. In: P. M. Bertinetto and M. Loporcaro (eds), Certamen Phonologicum: Papers from the 1987 Cortona Phonology Meeting (Torino: Rosenberg & Sellier) pp. 231–50. Vogel, I. and Kenesei, I. (1987) The interface between phonology and other components of grammar; the case of Hungarian. Phonology Yearbook, 4: 243–63. Vogel, I. and Kenesei, I. (1990) Syntax and semantics in phonology. In: S. Inkelas and D. Zec (eds), The Phonology–Syntax Connection (Chicago: Chicago University Press) pp. 339–63. Wacha, I. (1985) Beszéd: szituáció, szöveg és hangzás együttese a rádióban és a televízióban. [Speech: the unity of situation, text and sound on radio and television.] In: L. Grétsy (ed.), Nyelvészet és tömegkommunikáció [Linguistics and mass communication] (Budapest: Tömegkommunikációs Kutatóközpont) pp. 7–246.
Index abstract meaning (of intonation contours) see general meaning (of intonation contours) accent see pitch accent accented syllable see pitch-accented syllable accent phrase (AP), 60–1, 89, 191 afterthoughts, 123 alternative questions, 103, 105 ancillary tonal tier, 69 appended contour, 48–50, 52, 55, 56, 74–5, 96–7, 207, 209 Appendix Rule, 75 apposition, appositive, 104, 108, 113, 114, 121 association conventions (ACs), 66, 76–9 aspect, 144, 146 attitude, 23–4, 30, 91, 203, 204 attitudinal function of intonation, 22, 23–31 attitudinal variations, 63, 68, 117, 118, 120, 181, 202 autosegmental analysis/ representation of intonation, 65–86, 197 basic contours, 65–6, 70–1 “belittling imperatives”, 41–2 bottom pitch of the speaker, 8, 35, 48, 53, 69, 204 broad focus, 11, 140, 143 character contours, characters, 4, 7, 33–48, 56, 123, 127, 128, 207; see also end-falling, front-falling and sustained contours childish boasts, 46–7 Clash Deletion (CD), 136, 146–8, 151, 190, 194, 211, 212 clitics, 133, 137, 190, 191 Closing Rule, 71–2, 207 cognitive meaning, 24–8, 30
colon (Hammond), 179 comment (C), 10–12, 14, 24–5, 38, 57, 67, 88–9, 104, 130 communicative sentence type, 8, 24–8, 30, 91, 99, 204 Complementary Blocks (CBs), 102–12 complementary questions, 38, 67 compounds see ordinary compounds Compound Stress Rule (CSR), 134 conceptual meaning see cognitive meaning constitutive function of intonation, 22 configurational framework, 14 Contour Insertion, 138, 179, 180–90, 202, 209, 211, 214 contour replacement, 58, 99, 117–21 “contradiction contour” (Liberman and Sag), 23 coordination, 103, 104, 114, 121 Danish, 23 deaccentuation, 7, 137, 138, 140, 169, 178, 200; see also semantically/ pragmatically motivated deaccentuation, syntactically motivated deaccentuation “declarative imperative”, 45–6 declarative sentence, 139–40, 208 declination see downdrift derived contours, 66 derived intonation phrases, 181 descending contour (descent) 36–8, 44, 68, 72, 84, 105, 207; see also second-type descending contour descending-falling contour (descent-fall), 38–42, 72, 207 Designated Terminal Element (DTE), 13, 167 discourse, 5 discourse (discoursal) function of intonation, 31 224
Index downdrift, 58–9, 61–4, 68, 91 duration, 19, 203 Dutch, 166–7, 213 “eliminative stress” (Kálmán and Nádasdy), 144, 146 emphatic narrative see special narrative style emphatic sentence, 143 end-falling contours, 38–43, 61–2, 112, 197, 204, 205; see also rising–falling contour, monotone–fall contour, descending–falling contour English, 6, 21, 23, 26, 36, 38, 42, 44–5, 56, 57, 100, 130, 141, 145, 157, 160, 166, 169, 170, 177, 196, 197, 199, 201, 202, 206, 208, 209, 210, 211, 212, 214, 215 Equivalent Blocks (EBs), 100–2, 112–15 Exclamation Rule, 73 exclamations, 26, 36, 44, 197, 204 expletive verb phrases, 94 falling–rising contour (fall–rise), 23, 34–6, 72, 105, 110, 116, 117, 205, 207 Finnish, 201 Five-Syllable Constraint, 156 focus (F) position/constituent, 10–12, 24–5, 88–9, 101, 104, 139–40, 143, 208, 212 foot, 5, 33, 60–1, 89, 128, 132, 132–3, 190–1, 192 front-falling contours, 34–6, 42–3, 47, 61–2, 105, 112, 197, 204; see also falling–rising contour, fully falling contour, half falling contour fully falling contour (full fall), 8, 34–6, 49, 67, 72, 74–5, 84, 91, 105, 116, 117, 118, 120, 130, 180–1, 205, 207 function words, 136–8, 211; see also clitics Function Word Deaccentuation (FWDA), 140 functions of intonation, 22–31; see also attitudinal functions, grammatical functions of intonation
225
general meaning of intonation contours, 32, 205 German, 21, 23 given information, 23, 36, 144, 197 Gradual Descent see Rule of Gradual Descent Gradual Rise see Rule of Gradual Rise grammatical functions of intonation, 24–31 grid see metrical grid Grid Euphony Rules, 159–60; see also Hungarian Beat Addition, Hungarian Beat Deletion, Hungarian Beat Movement half falling contour (half fall), 9, 34–6, 49, 51, 57, 67–8, 70–1, 73, 85, 99, 105, 110, 112, 116, 117, 118, 121, 130, 180–1, 198, 205, 209 high monotone contour (high monotone), 36–8, 44, 68, 70–1, 105, 117, 118, 119, 120, 180–1, 209, 214 Highest-Ranking Sentence (HRS), 90, 91, 98, 99, 115, 121, 199, 208 Hungarian Beat Addition (BA), 159–65, 214 Hungarian Beat Deletion (BD), 159–65 Hungarian Beat Movement (BM), 160–5 Hungarian Rhythm Rule, 169, 214 Iambic Confirmation, 157, 166, 201, 214 Iambic Reversal, 157, 166, 201, 202, 212, 214, 215 Identical Tone Reduction Rule, 70–1 illocutionary acts, illocution, 8, 89, 91, 99, 204 imperatives, 36, 41, 205; see also “belittling imperatives” inorganic blocks, 93–7, 145, 146, 209; see also Non-Quoting Inorganic Blocks, Quoting Inorganic Blocks intensity, 5, 20, 33, 127, 128, 129, 130, 131 interjections, 210 intonation (definition), 19, 20, 196 intonation contours, 4, 7–8; see also melodic prosodemes, pitch patterns
226
Index
intonation language, non-tone language, 203 intonational lexicon, 4, 32, 50–2, 66, 197, 206 intonational notation/transcription, xi–xii, 9, 34, 128 intonation(al) phrase (IP), 4, 50, 52, 54–61, 85, 89, 90, 91, 136, 179, 180, 190, 191, 192, 193, 197, 198, 202, 206 intonation(al) phrase boundary (IP-boundary), 9, 34, 50, 53–4, 57, 90, 121, 181, 206 intonational theory, 2–3 Lexical Foot Erasing (LFE), 135, 210 lexical stratum of phonology, 7, 136, 137, 138, 201 lexical tone, 203 lexical word, 132–3, 140, 208 LH-Rule, 72–3 local meanings of intonational contours, 205 Logical Form (LF), 5, 60, 203 loudness, 19, 203 major stress, major-stressed syllable, 4, 7–9, 33, 127, 129, 130–1, 139, 140, 158, 168, 178, 180, 199–200; see also pitch accent, pitch accented syllable major tonosyntactic blocks, 90–7; see also Highest-Ranking Sentence, Non-Quoting Inorganic Block, Quoting Inorganic Block Maranungku, 179, 214 maximal syntactic category, 146, 147, 172 melodic prosodemes, 4, 20, 29, 197; see also intonation contours, pitch patterns melodic segmentation, 6–9, 85, 87–124, 198, 208 melody formation, 65–6, 71–5 melody rules, 101, 105, 106, 110, 111, 115–17 Metrical Adjunction Theory, 166–8 metrical grid, 131, 158–65, 166, 169, 201 metrical tree, 166
minor stress, minor-stressed syllable, 7, 33, 45, 47–8, 52–3, 127, 128, 129, 131, 137, 138, 178, 179, 180, 199, 202, 211 minor tonosyntactic blocks, minor blocks (MBs), 7–8, 97–121, 199–200; see also Complementary Blocks, Equivalent Blocks monotone-fall contour, 38–42, 70–1, 198, 207 narrow focus, 11, 143 new information, 139, 140, 207 Non-Quoting Inorganic Blocks (NQIOs), 93–6 non-restrictive relative clause (NRRC), 104, 108, 113, 121 non-rhythmically variable (non-RV) words/phrasal compounds, 151–2, 162, 163, 165, 169, 171 non-segmental devices see prosodic devices non-segmental phonation see prosody Nuclear Stress Rule (NSR), 14, 130, 131, 158–9, 161 nucleus (as part of the English tone-group), 56 numeric compounds, 152–3, 213 Obligatory Contour Principle (OCP), 70, 198, 207 Opening Rule, 72, 207 ordinary compounds, 133–6, 147, 211 P1 rules (Kaisse), 170–1 paralanguage, paralinguistic devices/ prosody, 22, 24, 25–6, 203 partial-degree constituents, 141 pause prosodemes, pause, 19–20, 48–50, 53–4, 57, 85, 91, 93, 99, 121, 203 Phonetic Form (PF), 5 phonetic realization rules, 66, 81–4; see also Rule of Gradual Descent, Rule of Gradual Rise phonological phrase, 60, 169–70, 212 phrasal compounds (PCs), 133, 149–57, 172, 210
Index Phrasal Compound Stress Rule (PCSR), 159, 172 phrasal stress, 138–46 pitch, 19, 203 pitch accent, pitch-accented syllable, 5, 7, 33, 57, 127–8, 131, 165, 168–9, 177, 199; see also major-stressed syllable pitch patterns, 19–20; see also intonation contours pitch range, 19–21, 24, 26 pitch register, register, 19–21, 22, 24, 26 Post-Focal Verb Deaccentuation, 140 postlexical stratum of phonology, 136, 137, 140, 146, 170–1, 178, 190, 191, 193, 201, 202, 214 postverbal (Pv) position/constituent, 10–12, 88–9, 103, 106, 107, 122, 144, 145, 146 precompiled lexical rules (Hayes), 170–1 preparatory part (of the Hungarian IP), 55–6, 60, 191 preparatory contour, 37, 52–4, 84, 94, 206 primary derivational rules see Opening Rule and Closing Rule primary stress, 128, 132–8, 210 profile A (Bolinger), 35, 118 profile AC (Bolinger), 35 profile B (Bolinger), 37, 118 profile CA (Bolinger), 38 prosodemes, 20 prosodic constituents units, 60, 89, 189, 208, 209 prosodic devices, 19 prosodic limits, 21 prosody, 19 qualitative reduction, 128, 210 quality, 19, 203 quantifier (Q) position/constituent, 10–11, 88–9, 212 quasi-function words, 141–2 question-word questions, 36, 40–1 quoting clause, 50, 94, 96–7; see also Quoting Inorganic Blocks
227
quotation carrying utterance (QuoUtt), 96–7, 206 Quoting Inorganic Blocks (QIOs), 50, 96–7 rearrangement rules, 66, 80–1, 207 reduced complement, 11, 140, 143, 212 register see pitch register Replacement of half fall by high monotone, 118–21 Replacement of half fall by fall–rise, 117 restrictive relative clause (RRC), 104, 121 Rhythmic Adjustment, 166 rhythmical variation of phrasal compounds, 131, 149–57, 200 rhythmical secondary stress, 178–80, 182–6, 201–2 Rhythmical Secondary Stress Assignment (RSSA), 60, 191, 193, 194 rhythmically variable (RV) words/ phrasal compounds, 150, 152–3, 158, 160, 163, 169, 171, 172 rising contour (rise), 6, 8–9, 36–8, 44, 67–8, 72, 95, 105, 122, 123, 207, 208, 214 rising–falling contour (rise–fall), 23, 29, 38–42, 44, 49, 52, 72, 85, 198, 205, 206, 207, 208 Rule of Gradual Descent, 81–2 Rule of Gradual Rise, 82–4 Russian, 6, 21, 23 scale (of the Hungarian IP), 54–6, 60–1, 191 scale contour, scalar contour, 55 second-type descending contour (second-type descent), 42–4, 68, 73, 84, 197, 207 secondary derivational rules see Appendix Rule, Exclamation Rule, Stylizing Rule secondary stress, 128, 132–8; see also rhythmical secondary stress semantically/pragmatically motivated deaccentuation, 144–6
228
Index
sense unit, 57–8 Sense Unit Condition on Intonational Phrasing (Selkirk), 181 sentence, 88–90, 123, 140 sentence qualifiers, 93, 145 shared knowledge see given information simplifying rules see Identical Tone Reduction Rule, LH-Rule Spanish, 215 special narrative style, 137, 138, 144, 209 speech acts, 204; see also illocutionary acts speech melody see intonation contour, pitch pattern Split Analysis (of Hungarian rhythmical variation), 170 S-structure, 5–6 statements, 26, 27–8, 36, 197, 204 stress 19 stress degrees, 33–4, 127–31 stress fixing, 6–7, 138–48, 200, 208 stress prosodemes, 20, 128–9; see also major stress, minor stress, zero stress stress timing, 201 Strict Layer Hypothesis (Selkirk), 59, 90, 209 strong local parsing (Hayes), 191 stylized falling contour (stylized fall), 44–7, 70, 73–4, 111, 117, 197, 202, 206, 215 stylizer, 69, 74 Stylizing Rule, 74 suprasegmental devices see prosodic devices suprasegmental phonation see prosody sustained contours, 36–8, 42–3, 52, 62, 105–7, 112, 116, 120, 197, 206, 209; see also descending contour, high monotone, rising contour, second-type descending contour syllabic tier, 65 syllabic weight, 180, 214 syllable, 60, 65, 89, 198, 206 syllable quantity see syllabic weight
syllable-timing, 201, 214 syntactic positions in the Hungarian sentence, 10–1, 88–9 syntactically motivated deaccentuation, 138–44 syntax-phonology mapping, 5–9, 196; see also melodic segmentation, stress fixing T model, 5–6, 41 Tamil, 22 taxonomic analysis of Hungarian intonation, 32–64, 204 tempo, 19, 21, 24, 26 terminal contour, 57, 181 terminal part (of the Hungarian IP), 54–6, 60–1, 85, 191 Textual Prominence Preservation Condition (TPPC) (Selkirk), 161 Text-to-Grid Alignment Rules (Selkirk), 158 thematic meaning, 24–8 tonal assimilation, 67 tonal tier, 65 tone languages, 203 tone-group, 56, 206 tones, 65, 69 tonetic stress marks, 34 tonic see nucleus (as part of the English tone-group) tonosyntactic block, 88, 90 tonosyntax, 208 topic (T) position/constituent, 10–12, 24–5, 38, 57, 67–8, 88–9, 103, 106, 107, 121, 130, 211 Transylvanian Hungarian, 208 Trochaic Confirmation, 150, 151–6, 160–2, 166, 169, 170–1, 172–6, 200–1, 214 Trochaic Reversal, 150, 151–6, 162–5, 166, 170–1, 172–6, 200–1, 212, 213 Trochaic Reversal Frame, 171 “ugyan”-compounds, 153 Ultimate Foot Erasing (UFE), 136 unaccented syllable, 128 unstressed syllable see zero-stressed syllable
Index upstep prosodemes, upstepping, 20, 63 utterance, 22, 60, 89–97, 123, 199
weak local parsing (Hayes), 193 word stress, 132–8 Word Stress Rule, 133, 138
verb (V) verb position/constituent, 10–12, 88–9, 139–40, 211, 212 vocatives, 45, 48–50, 94, 96 “vocative chant”, 45 voice quality variation, 19, 21, 24 volume, 19, 21, 24, 26
yes–no questions, 6, 23, 27–8, 29, 40–2, 44, 67–8, 145, 146, 197, 204, 205, 206
229
zero stress, zero-stressed syllable, 5, 33, 128, 129, 131, 199–200